Week 5 – 3D Scanning and Printing

Weekly Summary

This week I 3D printed for the first time (yeah!). First we characterized in the group work our printers at FabLab Kamakura, then I designed and printed (after some failures learnings) a hinge that connects 2 cubes together - and allows a 180ยบ rotation. This is a spiral development of my sketches from Week 2. We also used a 3D scanner to scan ourselves and a little buddha statute. My efforts to scan a really large statue did not come to fruition. We also did photogrammetry on the Great Buddha of Kamakura, and then 3D-printed and compared all out buddhas.

A 3D printed hinge with the pins in place, that can only be made additively.

Group Assignment

  • test the design rules for your 3D printer(s),

As we are 3 students at FabLab Kamakura, each could choose a printer and characterize it. Group Assignment Pageโ†—๏ธŽ is the shared results of these efforts, here is also my assessment of the Prusa Mini (Thank you to Rico for borrowing the printer!):

Image from the Prusa Site, forgot to take a nice image of our Prusa Mini

For the characterization of the FabLab Kamakura's 3D printers, we choose each a printer and prepared and printer the All-in-One Object from Kickstarter-Autodesk-3dโ†—๏ธŽ, local copy(612K).

This test geometry is testing the following, all with one object:

I downloaded the test geometry: https://github.com/kickstarter/kickstarter-autodesk-3d/blob/master/FDM-protocol/ksr_fdmtest_v4.stl

Using PrusaSlicer(on OSX 11.2.1), I imported the test geometry and choose the following settings:

After slicing, that gives us the following:

Here is the exported Gcode file: georg_prusa_test_ksr_fdmtest_v4_0.2mm_PLA_MINI_4h4m.gcode , (external link, 7.3M)

Turning on the Printer

Cleaning the Print Bed

With Isopropyl alcohol, because it evaporates fast and leaves the print bed nice and non-sticky. And it removes residual oil from the printing bed.

Auto-bed Levelling

One USP of the Prusa Mini is it auto-bed leveling. Before printing a sensor reading of the bed is taken, and the bed is adjusted accordingly, resulting in better print quality.

Importing the Data

Is done via USB Stick, the interface is in color, no touchscreen and a rotate-and-push dial (similar to Bento Lab)

Starting the Printing

I started the print process, here are some images during printing.

Printing Done!

Exactly 4 hours and 4 minutes later the printing was ready.

The main thing we choose this test geometry, was that it comes with it's own Assessment Protocol

Assessment

1. Dimensional Accuracy

I am measuring the x and y dimensions of the stack:

Target Measured X X Error Measured Y Y Error
25 24.8 0.2 24.8 0.2
20 19.7 0.3 19.8 0.2
15 14.9 0.1 14.9 0.1
10 10.0 0 10.0 0
5 5.0 0 4.9 0.1

According to the guidelines:

2. Fine Flow Control

Spines 40mm long - fully printed!, but lot's of stings: 2.5

3. Fine Negative Features

How many pins can be removed from holes, using only finger pressures, no tools? All can be removed: 5

4. Overhangs

5. Bridging

6. XY Resonance

Visually inspecting the ringing feature, for both X and Y:

7. Z-axis alignment

Is there a registrations effect with a period equal to the leadscrew? If so, 0, otherwise 2.5.

No, no registrations: Score 2.5

Summary of the Assessment

Check Points
Dimensional accuracy 4
Negative feature resolution 2.5
Positive feature resolution 5
Overhang 3
Bridging 3
XY ringing 2.5
Z-axis alignment 2.5
Total for Prusa Mini: 22.5

(But we don't know Slicing Parameters and Printing Times for these)

Conclusion

First time doing 3D Printing, very impressed with the results - and ease of use of the Prusa Mini. Mini+ just came out, would be nice to get one.


Individual Assignments

While the test object was printing - and all the printers were occupied - I finished the modeling of my subtractive piece. I am still in the process getting fluid with Fusion 360, modeling is taking longer than I expect it to.

3D Printing

I am spiraling up from the work from Week 2, where I sketched and prototyped a modular, configurable test tube stand. The unresolved issue, is how to connect the cubes together - and allow their 180ยบ movement.

This week, I am concentrating on the hinge mechanism.

Design Process in Fusion 360

Design in Fusion 360.

Slicing with PrusaSlicer

Slicing and Slice Settings are crucial, as they have a direct impact on the strength of the hinge.

Standard Settings, except I set the Vertical Shell Parameter to 4.
Compare Vertical Shell Thickness: 2, 3 and 4 Layers

Approach 1: Living Hinge (Failure)

I was curious if a living hinge could be printed with PLA. I modeled a test living hinge...

... printed it, ...

... and it immediately broke (faster than I could take a picture).

Apparently PLA is the wrong material for living hinges. My finial test tube stand will be a series of cubes connected with hinges. The hinges are the weak point, therefore I was also experimenting with being able to change the hinge only - and leave the cube intact. (I did not follow this avenue further, as the priority of the lab day was to get the assignment done. To be continued.)

Approach 2: Pin Hinge, Printed in Place

The second approach - which hopefully also qualifies a being made additively - is the Pin Hinge Approach. The challenge was to get the clearance of the hinge, cube and pin right, so it would not fuse - and also not allow to much travel.

Print 1

I could see after 10 minutes of printing, that I set the clearance parameter too low (0.1mm), the two cubes are fused together.

I stopped the print and started redesigning.

Failure.

Print 2

I increased the clearance to 0.2mm and printed again. This time it passed initial visual inspection, the cubes did not fuse immediately. But when the printer came to the hinge, I could see again the the clearance was not enough, the cubes would move, but they would probably break the hinge. I stopped printing and confirmed my suspicion.

Failure. Stop. Redesign.

Print 3

Ok, this is looking good...

And here it is, hot off the printer!

Hero Image

Hero Video

H.264, 720 ร— 570, 4.4M
Suge! and Sugoi! basically means "Great, it works!" in English (and in this context)

Conclusion

I did my first 3D prints today, 3D printing is working better than I expected, - and the printing time is longer than I expected. The bottleneck - for me - is still the design process.

Next Steps


3D Scanning

Testing the Scanner and 3D-Selfie

We used a Sense Version 1 3D Scanner from 3D Systems. We tried to scan a object on a turntable - we learned quickly, that the object can not be shiny. And that the black hair of my colleagues here in Kamakura is also too shiny!

We resorted quickly to scanning ourselves, also trying to do a 3D selfie scan:

Here is the resulting scan. Great fun, still looking for a practical application.

3D Scanning and Printing a Small Buddha

I used the _Sense Version 1 3D Scanner? again to scan a _Small Buddha.

Printing the Scan

Photogrammetry-ing the Great Buddha of Kamakura

A bit of history

Kamakura is famous for being the de-facto capital of Japan from 1185 to 1333 - and for the Great Buddha of Kamakura ("Daibutsu"), a monumental outdoor statute of Amitฤbha the Buddha of Immeasurable Light and Life. The statue is more than 13m high, measures more than 100 tons, when I visited it the first time - some year ago - it was still possible to get inside that statue. This is currently not possible. The statue was completed in 1243 and was originally house in a hall, which was destroyed by a storm, rebuild, destroyed again. In 1498 a tsunami destroyed the hall once more, since then the Great Buddha has been been outside.

Great Buddha of Kamakura

Attempt 1

I wanted to take images and photogrammetrically reconstruct the large, and well-known Buddha stature of Kamakura. I arrived at 15:46 at the place. The ticket both closed at 15:45, and although visitor were still inside, it was not possible to get. (I tried with all my charme!)

Security guard telling me in uncertain terms to leave.

Attempt 2

Together with Kai Naito we made another visiting attempt. This time we succeeded.

We walked from FabLab Kamakura to the Daibutsu, ca. 1.5km, signs showing the way.
Kai-san and me infront of the Daibutsu.

Together we made more than 250 pictures of the Daibutsu. Here are some examples:

Taking 250+ images the Daibutsu. (Some examples)

The total file-set is more than 500M, download it from here.

Photogrammetry

Photogrammetry is the reconstruction of 3D objects and scenes from a large number of overlapping photos.

Meshroom is a free and open-source 3D Reconstruction Software, that runs on either Windows or Linux, but in any case it needs an NVIDIA GPU - which my MacBook does not have.

For Photogrammetry on the Mac there are some tutorials that workaround the GPU challenge. This tutorial, aptly named 'Photogrammetry for Mac Users' is suggesting to use the following freely available applications:

Photogrammetry with NVIDIA

Kai-san also ran the picture-set through Meshroom on his Windows Laptop.

Kai's Meshroom
Kai's exported object.

Photogrammetry on the Mac

I also wanted to run Photogrammetry on my Mac - in order to compare if the NVIDI/Meshroom or Regard3D/Mac workflow yields the better results.

Adding Camera Info to Regard3D Database

Regard3D needs to know the sensor size and aperture of the camera with which the pictures were taken. Very likely your camera is NOT part of the build-in database. In that case open the /Applications/Regard3D.app/Contents/Resources/sensor_database.csv and add the camera info and sensor size.

Adding Camera Info to the Regard3D Database.

Import the Pictures & Compute Matches

Following the tutorial, I created a Picture Set, imported the images and clicked Compute Matches.

figcaption>Computing Matches

Triangulate

Once the matches are computed, the next step is to triangulate the matches.

Triangulated Matches.

Create Dense Pointcloud

Then, click Creating dense pointcloud... which will result in something like this:

Left: 200 pictures, Camera A, Right: 50 pictures, Camera B

Once the Pointcloud is created, you can either export the point cloud, or export the whole scence for use in MeshLab.

Cleaning in MeshLab

I opened the previously exported scene in MeshLab and remove stray points.

The limits of our Photogrammetry approach - and the massive size of the Great Buddha - became also clear: We simply did not have image from the top! (Challenge to future FabLab Kamakura Students: Fly a drone into the temple an take aerial photos!)

Simplification and Reconstruction

In any case, I followed the steps to turn the point cloud into a solid object.

  1. Filter > Normals, Curvatures and Orientation > Compute Normals for Point Set
  2. Filter > Remeshing, Simplification and Reconstruction, Surface Reconstruction: Screened Possion
From Pointcloud to Solid

Exporting as STL

Exporting as STL

Slicing

Left: Meshroom, right: Regret3D

Printing Budddhas

Printing Budddhas

Printed Buddhas

3 Printed Buddhas, Peacefully and Enlightened, Side by Side, 3.8M

Comparison

The clear winner is the NVIDI/Meshroom combo. Regard3D produces results, but they don't compare well with Meshroom. Also, a drone is needed to do Photogrammetry on large objects.

More Lab Session Images

Asako Okazaki and me working, Amazing Photos by Phanuwit โ€œRicoโ€ Kanthatham