Laser Scanning and Reverse Engineering Thread

[AgitatedPancake]

So some of you may know I've been getting into laser scanning and reverse engineering over the last year or so, but what I've come to experience is many people are hesitant not because they're not capable, but more because they're just not familiar with the capabilities. I wanted to make a thread I can continually bump just to talk about scanning and show people what they can realistically be capable of. Show you guys pictures of scanned components, the capabilities of the software paired with the actual scanner itself, and how it can be useful to just about everyone CAD tinkering (especially as computers get more powerful). For instance, did you know you can make assemblies in Solidworks strictly based on the scanned data without even having to make a solid model if you're just looking to use features of certain scans without having to reconstruct the whole component? It takes extra graphical computing power to not be slow just due to the extra amount of surfaces observed by the program, but can prove extremely valuable when it comes to time invested per assembly.

But I'll start with just some cool scanned 4x4 components, with followup posts showing some solid model creation and using raw scans in assemblies. Maybe some videos of the actual scanning process at a later date.

 

[AgitatedPancake]

Now here is where the software is worth its weight in gold, and makes the raw scan data something you can use for making accurate models, or even use without modeling at all. This is the single largest difference between awesome scan data and somewhat useless scan data. Good software (like VXElements) used with most quality scanners has the ability to observe the scanned mesh data, and create useable geometry with a heat map that shows the accuracy of the observed data.

So lets say this 05+ Super Duty knuckle gets scanned:

And you want to use it for an assembly, so you want to know where the unit bearing is, ball joint centerlines, caliper ears and tie rod hole. So within the scan software of those doing the scanning for you, they should have a system like this below. You choose a basic feature style you want (Plane, cylinder, sphere, cone, etc). For example in this picture I wanted a plane on the unit bearing mounting face, so I asked it to make a plane and selected the surface data for it to observe. It immediately turns it into a heat map of accuracy. As you can see in the legend to the right, anything in the green is within .003 of the plane, as it gets into the yellow and red it gets closer to .010 variation. Interesting note, it looks like this knuckle has been stretched/flexed a bit due to this heat map showing yellow at 12 oclock and 6 oclock.

After you have the mount face you want the unit bearing bore, so you choose to create a cylinder and do the same.

And soon enough you have all the features you want. You can also align it to a useful origin based on this geometry which is extremely handy. Then with a good software like this you can kick the whole mesh over to solidworks along with the seperate geometries.

So from this point on, you can decide if you would like to model the entire knuckle if your main focus is the knuckle, or you can use those new geometries to just drop it straight into an assembly with other axle components to create a complete articulating axle to double check clearances, cycle from steering lock to lock with a tire on to make sure tires clear body/shocks, or whatever you want to do. The below is a JK 44 scanned housing, with a solid model truss, and scanned knuckles along with scanned unit bearings all assembled together in a fully flexible setup that I can use to check clearances on everything exactly as it is in the vehicle. And the best part is I only modeled the items I had to create in the shop

 

[Provience]

awesome, thanks for sharing. this is certainly cool stuff and makes a bunch of stuff easier to design.

hell, i've never tried to use solidworks and still default to pencil and paper so i'm about 30 years behind the times on computer stuff, but it is awesome to see what can be done. certainly save a huge amount of material pairing this stuff with CNC 2D cutouts, laser or plasma table and CNC 3D stuff for milling and turning.

it's always seemed to be a whole system that lends itself to more of a production or higher volume environment instead of the one off stuff or a shop with less than 5 or 10 people

one of these days i'm going to get better at moving around in computer space, eventually "functional" isn't going to cut it

 

[JNHEscher]

I'm sure I'll be getting a scanner of some sort at some point. Fusion 360 is my go-to and it should support just about everything Solidworks does in this environment.

Keep this thread going. Awesome tech I certainly appreciate any education to be had from it.

 

[Yotota]

This is always rad to see for truck related parts. What are you using for the scanner itself, and then for a PC?

I used to do this on a larger scale... we scanned and modelled 10k SF - 100k SF manufacturing facilities for equipment retrofits and modifications. It saved so much time over the old manual field measurement trip with 2 engineers on site for 1-2 days. I could knock out a midsize plant in 3 hours by myself with more complete and accurate data. But then the processing and rendering could take HOURS, even on my fairly powerful drafting rig. And we would be dealing with huge 2+GB files...

 

[stan1688]

Awesome info!

I've looked into scanners a little and the good ones seem to quite a ways out of my price range. I look forward to what else you have to share.

 

[Tnsejed]

I'm an NX user and I've been wondering about these for awhile. Thanks for the writeup! Will you go into what you know about the actual scanners themselves? Along the lines of this have you done any testing via cmm or scope to figure out the tolerance of the scan itself versus the real part?

 

[Bebop]

Do you mind sharing some info on what tools you use ?

Also, do you have a library of models you scanned that you'd be willing to sell ?

Thanks

 

[Ghetto Fab.]

What are you using as a file format for you scans to get them to work in solidworks? Does your scanner output an .stl file?

Kevin

 

[AgitatedPancake]

Provience get on it man, it expands your capabilities to the next level, and is getting cheaper and cheaper to do.

JNHE, no problem at all man, and I look forward to keeping this thread up. The biggest thing I've encountered talking to people about laser scanning is just an overall lack of familiarity, but it is the future of aftermarket development IMO, and more and more people are going to be finding uses for it in the coming years.

Yotota that's badass, I've looked into the building-scale scanners quite a bit, and can absolutely see their value in the industrial world. I'm running a Creaform Handyscan 700, running on a laptop with an i7 and 64gb ram. File size can get big, but we're right at the edge of where peoples everyday computers are starting to become powerful enough to work around those larger file sizes so it's an interesting time to be doing this stuff.

Stan absolutely! The hardware is still expensive, but I'm working on the means to bring the file prices down to more reasonable levels for the masses, we'll see.

Tnsejed I'd love to compare against components measured with other methods, unfortunately do not have access to others to make that comparison. This hardware (Creaform Handyscan 700) is claimed accurate down to .0019", I assume that's under ideal conditions and don't claim every file is in that ballpark, but from what I've encountered so far, not far off.

Bebop, I'll absolutely be posting more about the equipment I use, and yes I do have a library of components that I've been expanding and willing to sell files for. I haven't upgraded to a vendor skull yet so I wasn't intending to do any direct advertising, but I'm planning on making that happen soon.

Ghetto, the VXElements software I scan directly into (what the above pictures are from) has the ability to export the mesh and reference features directly into Solidworks. I can then save as a .sldprt or any other format Solidworks is capable of. Catch is, other formats do not retain that reference geometry, so working with the .sldprt is the most ideal method

 

[Tnsejed]

Tnsejed I'd love to compare against components measured with other methods, unfortunately do not have access to others to make that comparison. This hardware (Creaform Handyscan 700) is claimed accurate down to .0019", I assume that's under ideal conditions and don't claim every file is in that ballpark, but from what I've encountered so far, not far off.

Thank you for your response!

 

[PAE]

Knew a cnc dude in town that before the camp fire made the speedo clusters from a laser scan of a perfect example of a factory piece, would love to do the same of a tach cluster base from a mid 1980s square body Chevrolet truck.

But the cost seems pretty steep up into the 4 digits area from talking to that guy.

Think its become any cheaper lately for the scanning at least ?

 

[AgitatedPancake]

Thank you for your response!

No prob man!

Knew a cnc dude in town that before the camp fire made the speedo clusters from a laser scan of a perfect example of a factory piece, would love to do the same of a tach cluster base from a mid 1980s square body Chevrolet truck.

But the cost seems pretty steep up into the 4 digits area from talking to that guy.

Think its become any cheaper lately for the scanning at least ?

Yep the price of scanning is coming down for sure. As far as scanning a factory cluster these days? Far below 4 figures

 

[AgitatedPancake]

And just some fuel for the imagination, what a combination of scans, a few readily available files off the internet and you're able to do comprehensive product development and can cycle check everything to the last detail. I can't tell you how many prototypes I haven't had to make and revise, how much time has been saved by having this model

 

[Eric]

Got some Chinese in ya apparently. LOL.

Cool stuff. Presumably, the scanner generates a point cloud which then gets tessellated into STL format? How big do the files get? Based on the resolution shown in these images there must be a gazillion triangles.

Years ago I worked on some projects converting CAT scan data into finite element models of bones and organs. There was never a need to convert the scan data into traditional CAD format. What is state-of-the-art for converting these models into a traditional CAD format (i.e. planes, cylinders, nurbs, etc)? You showed the heat maps in the scan software that can be used to locate datum geometry and then another picture of the scanned geometry in SWx. If I am understanding correctly, the datum geometry shown in SWx was created in the scan software although I guess the datums could have also been created within SWx (but without aid from the heat maps which AFAIK don't exist in SWX). What if you want to create relationships between the recovered features? For example, ensure that axis of the unit bearing bore is normal to the unit bearing mounting face? How do you go about converting a tessellated plane or cylinder into an analytical representation or converting some general tessellated surface into a spline representation?

Subscribed.

 

[Ghetto Fab.]

Whats the ball park costs for the hardware and software? This seems like something I need to get into. I do a lot of solidworks stuff anyways.

Kevin

 

[AgitatedPancake]

Got some Chinese in ya apparently. LOL.

Cool stuff. Presumably, the scanner generates a point cloud which then gets tessellated into STL format? How big do the files get? Based on the resolution shown in these images there must be a gazillion triangles.

Years ago I worked on some projects converting CAT scan data into finite element models of bones and organs. There was never a need to convert the scan data into traditional CAD format. What is state-of-the-art for converting these models into a traditional CAD format (i.e. planes, cylinders, nurbs, etc)? You showed the heat maps in the scan software that can be used to locate datum geometry and then another picture of the scanned geometry in SWx. If I am understanding correctly, the datum geometry shown in SWx was created in the scan software although I guess the datums could have also been created within SWx (but without aid from the heat maps which AFAIK don't exist in SWX). What if you want to create relationships between the recovered features? For example, ensure that axis of the unit bearing bore is normal to the unit bearing mounting face? How do you go about converting a tessellated plane or cylinder into an analytical representation or converting some general tessellated surface into a spline representation?

Subscribed.

Haha honestly, the past 5 or so years the capabilities of the everyday home user has been absolutely going through the roof. With extremely powerful computers becoming normal, 3d printers ending up in everyones home for tinkering, I just see accurate (and easily attainable) scanned data as the next piece of the puzzle that's going to allow everyone to become more and more capable.

The scanning software (VXElements) does generate a point cloud and automatically tessellates it into a mesh on the fly. But it's stored in a proprietary .CSF format, so an STL is just one available export format. I can change the mesh resolution on the fly to manipulate file size as desired depending on the accuracy I'm looking for. For example the complete scan of the XJ itself I have sectioned into 3 sections, each at 15gb in the resolution I have chosen (so 45gb total). When I have them combined, I have it dialed back so the complete scan is only 15gb. But by the time it hits solidworks, the complete mesh is less than 1gb so it's very workable.

Creating a solid model from the mesh and reference geometry is where it becomes a bit of an art form as there are many ways to accomplish it. But the more references you have such as planes to work off of, the less time it takes to recreate the features in geometric form. I create all of that reference geometry in the scanning software with the assistance of those heat maps. Solidworks has no easy way of creating those references (yet) that I'm aware of, so it's nice to bring them straight in with the mesh. You ask about relations between recovered features, that is a huge part of what the geometry can do. For instance in my above examples of the unit bearing mount face and bore, I can create the mount face plane, and then as I'm creating the cylindrical feature for the bore, I can tell it that it is intended to be perpendicular to the mount plane. Or in other instances - planes parallel to other planes, cylinders sharing a single axis (or just having the axes parallel), or various other valuable assigned relationships. Another feature I haven't discussed yet, is you can have it create a surface off the mesh, to truly follow organic shapes and curves if needed.

Ghetto - you can get into scanning for somewhere around $4k, and you can go north of $40k. My equipment is in the mid-upper range, but not at the top.

 

[Yotota]

Ghetto - you can get into scanning for somewhere around $4k, and you can go north of $40k. My equipment is in the mid-upper range, but not at the top.

While the units we used were $50k-ish IIRC, we decided not to buy one and just contracted with a local surveying equipment supplier to rent one whenever we needed to use it, as they were an official dealer and rental supplier for Faro (scanner company). The rental cost was passed on to the client in the survey fee which was usually comped once they placed an order for our equipment.

So for handheld scanners, definitely look into rentals in your area.

 

[AgitatedPancake]

If you're comfortable with the software, rentals are not a half bad idea. I wonder how they deal with software licensing on rentals, is that something you have experience with Yotota?

One thing I had a mild understanding of before getting into this, and I've become faaaaar more familiar with as I've experienced more and more. You have to consider what you're scanning. You are not scanning a perfect item, you are just scanning something that was within tolerance when built, and has usually seen decades of use/abuse/stress. So there is an art in making the best out of the item you're observing. If it's a spline or a bearing bore - scan and measure it, then cross reference it across the Machinery Handbooks spec, or the manufacturers spec. Such as Timken. You measure bearing bores, you note the part number of the bearing used, cross reference it across their catalog and they actually denote the +/- tolerance they expect for that bearing. So you can confirm what you measured is within spec, but you're not actually using the scan as the final spec, because there is a book number you should follow. Or when you're scanning castings. Sometimes castings like knuckles or steering boxes aren't clamped perfectly square in their vices before final machining, so you have machined features not perfectly centered within cast features. This is where it becomes an art form, and open to quite a bit of artistic discretion. If you are reverse engineering a model out of these scans, are you trying to recreate the exact (but imperfect) example you have in front of you, or trying to interpret and recreate the original "perfectly" dimensioned model the manufacturer would use? Usually I choose to try to create the manufacturers original intended dimensions which are open to a certain tolerance, even though the model I make may not perfectly align with the scanned data. But that is a crucial discussion to have while approaching laser scanning, as it can make a difference on available chassis clearances etc due to manufacturing imperfections.

For example this steering box I did for Joe Thompson building the UFO cars here a couple weeks ago. The actual primary piston bore is not perfectly centered in the casting. I chose to recreate my interpretation of the "intended" dimensions, even though they vary from the actual dimensions of the one box I measured which seemed to be at the extreme limit of tolerances. In the below pictures the light gray is the scanned data, the dark gray is the solid model. Then you'll see them overlapped, and you'll notice the light gray cast extending beyond the solid model on one side due to an offset just created by how it was clamped into its final machining fixture. It doesn't cause any problems in application, but has to be interpreted while measuring.

 

[Yotota]

If you're comfortable with the software, rentals are not a half bad idea. I wonder how they deal with software licensing on rentals, is that something you have experience with Yotota?

Oh yeah, the Faro Scene software was another critical component and as per usual, proprietary to the scanner. Our rental agreement included a USB dongle with a license key to run that software and the rate to keep it an extra few days after each job wasn't too bad. Much better than investing in a single seat of the software at $5-6k...

 

[ParadisePWoffrd]

I know you have and are talking about scanners, but has anyone tried photogrammetry? For automotive or large parts it seems like it might not work, but seems like a lower cost of entry.

 

[Yotota]

I know you have and are talking about scanners, but has anyone tried photogrammetry? For automotive or large parts it seems like it might not work, but seems like a lower cost of entry.

Yep, from what I've seen and tested it's feasible for obtaining some rough data on a larger scale, but not accurate enough for machining/manufacturing purposes.

It's a cool tool for something like rough building layouts on a piece of property, or rough room layouts and areas.

 

[Eric642]

Nice work! How much does it cost to have something scanned? I would be scaning EVERYTHING if I had that toy available lol

 

[Austin]

Damn those are some sick scans!

 

[gmcxt]

I was today years old when I discover this scifi tech on IG. First thing i do is search it here and boom my head exploded.
This is exciting.

Camburg Engineering on Instagram: "Any RAM TRX owners out there? Scanning RAM TRX parts so we can start designing high end parts for this platform. Billet upper arms, billet rear links on the way. #ramtrx #ram #trx"

2,843 likes, 44 comments - camburgracingDecember 16, 2021 on : "Any RAM TRX owners out there? Scanning RAM TRX parts so we can start designing high end parts for this platform. Billet upper arms, bill..."

www.instagram.com

 

[Gdogg]

This technology to be able to scan and duplicate is priceless to anybody who builds anything, just ask Chi-Na, lol

With this it'll take your build to the next level and it'll be cheaper and faster.

 

[Bray D]

Bumping this thread to keep it fresh. I'm a design engineer at a company that makes aftermarket parts for the mining industry. We've been using scanning technology for nearly a decade. I've used multiple arm based scanners and target based handheld scanners, Handyscan included. All of that is to say that this is my jam and I'm stoked to see AP using it 'recreationally'.

I very much want to get my hands on my own hardware so I can do similar things at the home shop, but it's still cost prohibitive at the moment.