Week 05 - 3D Scanning and Printing

Group Assignment:

• Test the design rules for your 3D printer(s)
• Document your work on the group work page and reflect on your individual page what you learned about characteristics of your printer(s)

Please refer to the group page to read about our group assignment, where I was responsible for the Prusa MK4S.

Individual Assignment:

• Design and 3D print an object (small, few cm3, limited by printer time) that could not be easily made subtractively
• 3D scan an object (and optionally print it)

Learnings from the group assignemnt

3D Scanning

I decided to follow the recommendation of my supervisor to use Meshroom for photogrammetry as we have no 3D scanner in our lab. To get familiar with the process I watched this tutorial.

My first object is a small 12V LED-working light. It has an aluminium housing, glass lense and mounting points.

First I decided to cover the glass lense with a piece of paper to minimize reflection and found a bolt that I can hold the light with.

Then I set up the photo studio and got my camera.

And then it was time to go several circles around the object, shoot a lot of photos from many different angles and transfer them to the PC.

Here are the 39 images loaded in Meshroom and the processing has just started.

After 30 minutes the process stopped at the "Depth Map". I remebered in the tutorial the Nvidia GPU was mentioned, but it didn't come into my mind, that the new and powerful laptop I have might not be equipped with one. (It has an Intel Iris Xe GPU)

I found this help page where this issue is discussed. Now, there seems to be a version (ZLUDA) that works for AMD GPUs, but that doesn't help me.

Because I don't want to use a cloud service, I might have to live with the draft mesh, if I don't find another computer with Nvidia GPU.

In this tutorial the use of Meshroom CL was recommended and I tried it, but it failed in the TexturingCL Node. Maybe my images aren't good enough?

So I went back to the initial Meshroom 2023.3.0 and activated the DraftMesh Pipeline. To do so, go to File - New Pipeline - Photogrammetry Draft.

Then I ran the new pipeline, but it failed at the Meshing node.

Error

Failed to estimate space from SfM: The space bounding box is too small.

That sounds, like the images aren't good enough or the background is causing trouble? I tried again taking extra care when photographing and also took more images this time, so I was confident, that it would work, because it looked so easy on the youtube tutorials.

It finished the pipeline this time, but the result is horrible. From the camera view you can see, that it didn't detect the position of the camera except on one side of the object. I don't understand why.

My best idea at this point was to try a larger outdoor object, so I went to sculpture in town and repeated the process.

This time it worked better and the pipeline finished with something that looked more promising in the preview.

So let's take the mesh into Meshmixer and see if we can get a useful model out of it.

After a while of transforming, scaling, cutting and improving I got a low detail, but okay model.

I also printed the result for fun.

However, I wanted to try the proper pipeline with an Nvidia GPU and see if the result differs. I went to the Fablab, where my instructor gave me access to a PC with Nvidia P400 GPU.

I copied all the files to a memory stick and run the photogrammetry pipeline but it stopped for no obvious reason in the "Feature Extraction" node.

We tried to figure out what was going on and ended up excecuting the following steps:

• establish internet connection
• copy Meshroom and photos to c-drive
• save the project in the local folder

Then, miraculously it worked and the pipeline finished for the LED-light photo set, but the result is quite similar and no traces of the actual object are visible.

Next I processed the statue again and in the meantime my instructor suggested to try covering the light with baby powder and make a new set of photos, which we did.

3D Printing

As I'm quite familiar with FDM printers I decided to design a part that includes joints and will be able to move in the joints after printing.

A nice mechanism I discovered recently is the Harris joint.

It is described as a high torque and high articulation angle continious-velocity joint.

Designing

I did the design in Fusion360 and made the clearance in the joints parametric so I could adjust it later if needed. The part itself has a diameter of 50 mm.

Here is a screenshot of the design:

(I didn't invest the time to make it a proper assembly for being able to make mechanical joints in Fusion and simulate the movement, but it would have been a nice addon. In this case it is only a multi-body component.)

Slicing

I decided to use the Prusa MK4S that is in my office and black PETG.

The slicing software is PrusaSlicer 2.9.0.

Print Settings

Nozzle 0,4 mm; Layer height 0,2 mm; Infill 50%, Organic style support

Print duration: 8h 30min Filament used: 81,1 g

Printing

The printing was uneventful.

Here is the piece while the print is in progress.

After Printing

After printing I just had to pop it off the build plate and clean of the support. This is best done using a sharp screw driver and some wire cutters.

As you can see the joints worked, but they were quite hard to move and the part wouldn't move as easy as I expected. Also the wall thickness at some joints was a bit to thin, so I cracked the plastic, when trying to break the joints loose.

I'll make another iteration with more clearance (increased from 0,2 mm to 0,3 mm) and reinforced joints.

Now that is much better. I'm quite happy with the result.

The files for this week can be found here.