Budget 3D scanner vs. Metrology lab? Scanning a Tesla Model 3 gearbox with a Shining 3D EinScan HX

Welcome. This is my first 'real' post on this blog, with the previous post essentially just being a "hello world". In this post I'm going to discuss capturing 3D geometry and interpreting that data to do something meaningful with it, in this case; make some gears. 

I'm making a new gearset to replace the gears in the Tesla Model 3 rear "drive unit", the "980" drive unit to be specific. Why, you didn't ask? Well, I'm using these motors for purposes other than automotive and require the ability to manufacture my own precision gearsets to achieve whatever gear reduction ratio I desire. As shipped from the factory in a Model 3 the gear reduction is ~9:1, and my present application desires closer to 1.5:1. A few years ago we did this same process for the Model S Sport drive unit, but as manufacturing has halted for that product and every high end (S3XY) Tesla coming off of the production line is using a Model 3 PMSRM-derived motor, it makes sense to make this Model 3 motor my bitch. 

I've tried several 3D geometry capture methods over the years; line laser and structured light scanning with David 3D Laser Scanner (later acquired by HP), 3D scanning with the NEXTENGINE scanner, I've used MicroScribe CMM arms, photogrammetry with Autodesk, and paid metrology services. 

Haven Metrology Model

For this project we contracted Haven Metrology to perform a scan of the Tesla Model 3 gearbox half shell and provide a model to work with. Haven also did the scan for our Model S gearset, several units (and years) later things are still going well with those gears. Haven provided an STL with the bearing pocket cylinders modeled and a plane on the face of the mating surface where this part bolts to the rest of the assembly. This is what I will use as the "holy grail" measurement for which to base all gear designs from as we have done previously with great success.

EinScan HX

Concurrent to having the 3D scan done with Haven, we also reached out to Shining 3D about their EinScan HX series 3D Scanner.  This hybrid (structured light and blue laser) scanner's features look comparable to much more expensive scanner options. There are several scanner options in the sub $15K price range, maybe that's a topic for another post should I have an opportunity to evaluate other scanners at some point (please let me know if you want a shipping address to send me one ;). For now, this is what we have to work with.

I was put in touch with our local Shining 3D Rep, Grant Michel, from Wurth Additive, who has previously worked at Shining 3D in China. Grant has a ton of experience with not only the Shining 3D products but photogrammetry data capture and metrology in general. Grant drove over two hours to do an on-site demo of the EinScan HX, the data captured from that demo is what I'll be using in this blog post.

This project offers a great opportunity to compare the output from the EinScan HX to a paid metrology service, as we are only going to be doing more scans in the near future and having a kickass in-house 3D scanning solution has been on my equipment lust list for longer than I have logs. I've drooled over many various handheld or tripod/turntable scanner solutions, Faro and Hexagon arms with probes and lasers..oh my....🤤.... Until fairly recently most 3D scanners under $25-35K have been majorly consumer/prosumer rather than professional, and certainly not "metrology", grade product. Another post could be dedicated to what is "metrology grade", but today we'll summarize it as "the good shit", made of unobtanium, with traceable regular calibrations and exceptional repeatable accuracy. For the vast majority of my needs metrology grade is not required, however machining these gears properly the first time justifies paying the metrology lab fee.

Lets look at a couple ways to obtain measurement data from scan data. While the Haven model is super simple because they have already modeled the cylinders for me to measure in CAD, we do not have that luxury when performing scans yourself. When you scan something the output of that scan effort is a mesh file containing millions of data points in 3D space. You, the human interpreting that data, have to make a determination as to what points are relevant and what points are not, and where to draw geometry representing your scan which you may take measurements from.  Depending on the software packages available to you, this could be an easy task or incredibly challenging. Take for example these cylinder walls. I need to know "exactly" (or the "real") spacing between these bearing pockets in order to make a precision gearset. 

<rant> 

What is real? Real is whatever the hell you say is real within the specified tolerance. In most any situation where you are requiring parts to be machined you will be required to deal with "tolerance". Tolerances are fantastic, as it helps the machinist know how close to the specified dimensions that machining operation can be for the part to be considered acceptable (as in bad things don't happen if your part is inaccurate within plus or minus a specified amount. The more precise your tolerance, the more expensive that process becomes). You can dive deeper into tolerances here. For this particular gearset, my acceptable machining tolerances are very tight, so having accurate data to base the CAD dimensions for part design from is pretty important.

</rant>

Scan data from EinScan HX

Scan data from EinScan HX

 

If I only have Fusion360 at my availability I would start by creating a section analysis so I can work perpendicular to the cylinder opening, allowing me to locate an offset plane below the surface of the cylinder opening.  I may do several of these at varying depths from the top surface and observe the result. Is it drafted (meaning this is actually a cone rather than a cylinder)? If not, I may average the results to have more confidence in my measurement.

 

I then use Fusion's "Create Mesh Section Sketch" tool to project sketch entities where the plane intersects the mesh.

I can then use "Fit Curves to Mesh Section" with the "circle" type selected and give the inner circle a click. This gives me geometry that I can work with to sketch points and eventually measure against another point. I can then repeat this process for the other two cylinders on a different plane (as they are co-located at a different Z height).

Once we have circles, we can easily measure between the center points of those circles to determine our distance required for the gears to properly mesh. Here are the results I obtained using this method:

Now lets look at how to accomplish this with a popular software package designed to aid in mesh analysis, Geomagic (now 3DSystems). I'll be using the trial version of Geomagic Essentials to help measure the cylinders.

With Geomagic, I can select some of the inner face of the cylinder to use as a reference and use the "best fit" cylinder option, repeating this for each cylinder. I can then export the cylinders to Fusion 360 or any other CAD package to inspect. 

They come in as unstitched surfaces, so I have to tell fusion to stitch the four faces for each cylinder together to get a solid body. Once I have that, I can take my measurements of the center points. I did this process five times for this test. Five exports from Geomagic, five sets of cylinders to stitch in fusion, five drawings generated and data logged into a spreadsheet to easily average. 

Results!

As you can see, the results are crazy impressive for a reasonably affordable ($10K) professional scanner. While I will still be using the data from Haven Metrology for making the gears, I can feel pretty confident using the EinScan HX for most applications with tolerances that are +/- .004" (.1mm) or higher. 

I think I can quit here, I may append more to this or follow up at a later point. I need to get back to work :)

Thanks for reading!

Kaleb

Find me exclusively on Twitter! @t3nable