WEEK 5 | 3D Scanning & Printing
3D Modeling - Voronoi on Grasshopper
I want to use this week's assignment to make progress on my final project. I'll try to design and print a first version of the "electronic tower", which is supposed to contain the battery and the PCB board. I've also discovered the parametric function on Grasshopper called Voronoï and I really want to play with it. So the challenge of the week will be to design and print that "electronic tower" using Voronoï on Grasshopper and Rhino.
First, I explored a lot of tutorials to figure out which ones would be the most easily adapted to what I want to do. A few were useless because they used old Grasshopper functions. I decided to follow this one in order to project a Voronoï design on a surface. The idea is to create a Voronoï surface that I could duplicate and rotate in order to make a "volcano shape". I needed the function "Remove the duplicate lines" so I loaded the Kangaroo library for Grasshopper. I followed this tutorial carefully but it didn't work. It might be because of my surface slope. Bry helped me to find another way to deal with Voronoï on Grasshopper:
To make it easier, we used Grasshopper's predefined cone. Then we added random points on it. Thanks to a slider, I can control the number of points.
The next step was to find a way to transform the lines in tubes. I tried to use the function "pipe" but it only works with planar faces and the result was a cube instead of cone.
Download here the Grasshopper file.
Bry helped me once again to find a way to make pipes on my cone shape. But then I had another issue to deal with: tubes stopped at the end of each segment and it created holes, A LOT of holes.
I've tried to deal with this problem on Netfabb but without the Pro version, I was just able to close holes as you can see on the first screenshot. I wasn't satisfied so I switched to Meshlab and used the "Poisson" surface reconstruction function. The result was way better.
I had to go back on Rhino to finalize my shape. I checked on Netfabb before printing and everything was GREEN. So I was confident when I started to print.
3D printing
Fab Lab Wgtn uses the 3D Printer "Up" which has its own software. Here are the different steps to follow:
- Initialize: connect computer to printer
- Start preheat (because it takes a while)
Advice: don't print files from your pin drive
- Import your file
- Select the quality of your print (normal/fine/fast)
- Set your "Preferences" such as the "Z-resolution" (thickness of your plastic layers: thiner -> better -> longer)
- And print!
I left my first print to grab some food and when I came back... I discovered an unexpected creation! The 3D printer was confused because I didn't merge the bottom of my cone with the disc below. The machine tried to print the part of the cone which was inside the disc below.
Z-Resolution: 0,20mm
Part: Angle: 45 Deg & Surface: 4 layers
Support: Dense: 3 layers, Space: 8 lines, Area: Only Base, Angle: 30 Deg.
3D scanning
At the beginning of the week, Craig took time to introduce the different 3D scanning methods that we could use.
Roland:
This machine uses a laser to scan. It has two eyes which move up and down. It also uses a bed to make the object turn. So the object has to be stable and covered with an non-reflective material (I you need, use matt paint). The software, LPXEZStudio, always does a preview scan. It says how long the real scan will take depending on the parameters you choose. To say how precise you want to scan, change the square size: more little are the squares, more precise is your scan.
Using a Kinect:
They use a free software called Skanect. How does the Kinect scan work ? Turn around the object with the device, don't go to fast, don't be too close (1 meter away). You also have to check the light and try to stabilise your object as much as possible. Using a tripod helps a lot (apparently). The Kinect detects colors. The problem is when you export the file as an .stl, the quality of the mesh is reduced. The initial file has too much informations and is too heavy.
Photogrammetry with a camera:
Take a lot of pictures turning around the object and go higher and higher every round. Then you can import your photos on "123D Cad". This could be used to transform your pictures into a 3D model.
I wanted to scan a leaf. First, I tried to do it with the Kinect. I used nylon wire to stabilise it. The Kinect wasn't detecting the object. So I tried to:
- Change the background (many times)
- Turn off the light
- Change the disposition of the leaf.
Nothing worked! Finally, Stuart told me that it might be because vegetation absorbs light in a different way. It may disturb the Kinect. So I decided to paint it to see what would happen. But it didn't work either. The only answer was: my leaf was too small to be detected.
x
I switched to the photogrammetry method and took a lot of pictures. I used PhotoScan to generate the 3D model. I "Added" and "Selected" the pictures, then I "Aligned" my photos and set the "Accuracy" (high) and the pair preselection (generic).
Once your model is created, you can generate the mesh. This is my result. Now I need to clean it and fix it on a 3D software.
The photogrammetry result wasn't very successful so I decided to try scanning with the Roland Scanner. I set the machine to do a very precise scan: width:0.2 & height:0.2. It took around 7 hours so I did it over night. But when I came back, I realised that the machine didn't detect the leaf. Conclusion: it's very complicated to scan an object as thin, as small and as complex as a leaf with one of these three methods.
Download the .stl scan file