Skip to content

5. 3D Printing and Scanning

Global Class — (Adrian Bowyer)

This week’s global lecture by Adrian Bowyer introduced us to the world of 3D printing by one of its pioneers.

While the content of the presentation was very valuable, I would have preferred hearing anecdotes, personal experiences, or stories from his involvement in the development of additive manufacturing to make the session feel more engaging and dynamic.

Global

Nevertheless, the lecture introduced the fundamental differences between additive and subtractive manufacturing, particularly in terms of geometric capability, computation, and material efficiency.

In subtractive processes such as milling:

  • Material is removed from a solid block
  • Cutting tools must physically reach all surfaces
  • Internal enclosed geometries cannot be accessed
  • Multi-axis machining increases computational complexity
  • Increased geometric complexity increases machining time
  • Parts are often manufactured separately and later assembled
  • Significant material waste is generated

As a result, subtractive fabrication is constrained by tool accessibility, making it difficult or impossible to produce internal features such as enclosed cavities or lattice structures.

Additive manufacturing instead builds parts layer by layer, depositing material only where needed. This allows for:

  • fabrication of internal cavities
  • enclosed channels
  • nested geometries
  • internal lattice structures

These can be produced in a single operation and would not be manufacturable through subtractive means due to lack of internal tool access.

Additive Manufacturing Technologies

The lecture introduced Fused Filament Fabrication (FFF), a material extrusion process in which thermoplastic filament is heated and deposited through a computer-controlled nozzle.

Thermoplastics:

  • soften when heated
  • solidify upon cooling
  • can be reheated and reused

Thermoset Processes

Other additive processes rely on chemical reactions rather than cooling to harden materials.

Stereolithography (SLA) uses photopolymer resins that cure under light exposure, either through vector-based laser curing (SLA) or pixel-based layer projection methods such as Digital Light Processing (DLP) and Masked SLA (MSLA).

Because these systems rely on uncured photopolymers, appropriate safety precautions are required, including the use of gloves and UV-protective glasses.

Post-processing typically involves print removal, isopropyl alcohol rinsing, ultrasonic cleaning, drying, and final curing in a UV chamber to complete the polymerization process and stabilize the part.

The lecture helped me build a basic understanding of how 3D printing differs from other fabrication methods and guided my the design decisions to explored for the rest of the week.

Local Class

SLA Workflow

Observed post-processing workflow for resin-based printing included:

  • Print removal from build plate
  • External cleaning using isopropyl alcohol
  • Ultrasonic bath cleaning
  • Rinse and drying
  • Final UV curing in curing chamber

This final curing step ensures full polymerization of the thermoset material and stabilizes the printed part.

Catrina

Global

Global

Global

Global

3D Scanning

This exercise introduced the workflow for:

  • capturing physical geometry
  • reconstructing digital mesh data
  • exporting STL geometry for fabrication

Scanning technologies discussed included systems derived from:

  • personal mobile hardware
  • depth sensing technologies
  • LiDAR systems used in autonomous vehicles

Scan

Clay workshop

As part of the local class, a clay paste modeling session was conducted.

Clay

Global

RAGFAB Setup & Test

During the week I took a detour as part of AI recitations to explore possibilites of using it in Fablab context. This is the documentation for that stage paralell to 3D week to test if it could expand capabilities.

Based on Cesar’s RAGLab + Amira’s AI + Design presentations,

Objective

Test RAGFAB + AI Design workflow to:

  • Run models locally
  • Use Fab Academy documentation as grounded dataset
  • Support fabrication-aware CAD decision making for ASFALT hardware

Setup Performed

  • Installed GPT4All

  • Downloaded:

  • Mistral Instruct

  • DeepSeek

  • Nomic Embed v1.5

  • Added OpenAI API key (remote provider option)

  • Indexed:

  • FabRAG webcontents_raw folder via LocalDocs

  • 3D printing global lecture transcipts
  • Resources from Fablab links like 3D Printing Handbook and Fablab Kerala guide

Global

Intended Test Use Case

Instead of prompt engineering, the idea was to create a context engineering, to teste system based on a specification sheet to generate CAD instructions.

Task was to generate Rhino + Grasshopper modeling workflow for:

Octagonal wa-style handle (Load-bearing, thermally isolated, additive-only geometry)

Based on:

  • locked dimensional constraints
  • FFF/FDM print process
  • internal lattice
  • support-free requirement

Expected Output

Should include:

  • Polygon(8) loft (not circular sections)
  • 16 mm ferrule as split plane
  • load path through Ø12 mm steel rod
  • Volumetric internal lattice (21–81 mm zone)
  • XY-aligned print orientation (anisotropy-aware)
  • Overhang <45°
  • Additive-only internal geometry
  • No trapped supports
  • Subtractive tool-access limitation

Observed Output

Major CAD + fabrication errors:

  • Used circular loft → lost octagonal grip geometry
  • Generated surface pattern instead of volumetric lattice
  • Reversed FFF anisotropy logic
  • Suggested upright orientation (layer-splitting risk)
  • Proposed supports instead of support-free geometry

Interpretation, Outcome and Conclusion

Based on Cesar’s RAGLab work and Amira’s AI + Design lecture, it became clear that setting up a local RAG workflow involves more than indexing a dataset and expecting magical results to appear.

In this case, for retrieval to be meaningful, the system needs:

  • curated and task-relevant examples “few shots”
  • domain-specific prompting, skills, tools, etc
  • structured ingestion of fabrication knowledge

From a design perspective, I could not create the right AI-assited dataset and workflow to generate CAD isntructions or manufacturing tasks. The output failed dramatically and future iterations ended up in hallucinated workflows.

Although the local embedding and retrieval pipeline was technically functional, it did not produce domain-grounded CAD reasoning. I just tried the first part of the RAG project, so additional step would definitely create better results.

In conclusion, running local models with FabRAG enables context engineering and private retrieval of documentation that can be valuable for other purposes in the future, it wasn’t suitable for additive design tasks, so I had to go back and design it myself!!

Individual Assignment

WA handle for Final Project (Asfalt)

The objective of this assignment was to design a component that:

  • demonstrates additive-only geometry
  • follows ergonomic principles
  • cannot be manufactured subtractively

A handle inspired by traditional Japanese Wa-bocho knife handles was selected as a design reference due to:

  • defined ergonomic geometry
  • prismatic multi-face grip
  • tapered longitudinal profile

Global

This was also intended as a nod to Japanese craftsmanship in a cooking environment. I personally own knives with this type of handle and always appreciate the grip, ergonomics, and overall tactile feel during use.

Global

Inspiration

The Wa handle is traditionally used in Japanese knife-making and features an octagonal cross-section that:

  • improves grip stability
  • prevents rotational slip
  • distributes pressure across flat surfaces
  • reduces fatigue during repeated use

These properties are relevant for the Asfalt radiant hood mechanism, where the handle is intended to be repeatedly pushed and pulled during service operations in a high-temperature cooking environment.

Reference inspiration for the ergonomic geometry was taken from traditional Wa knife handles.

Additionally, a 3D printed Wa-style handle produced by:

printsThatFunction

was used as a reference for understanding how this traditional form factor could be adapted into an additively manufactured component.

Global

Design Process

The design process began by translating the ergonomic logic of the Wa handle into a geometry suitable for additive manufacturing. Unlike traditional wooden Wa handles, which are typically machined or manually shaped, this version was modeled digitally in order to take advantage of internal structures that cannot be produced using subtractive fabrication techniques.

Using Fusion 360, the external geometry of the handle was constructed by lofting between two octagonal profiles:

  • Rear profile: equipment-scale ergonomic grip
  • Front profile: reduced diameter for tapered feel
  • Total length: 180 mm

The octagonal cross-section was selected in order to:

  • improve tactile orientation
  • prevent rotational slip during handling
  • distribute load across flat grip surfaces

Both profiles were placed 180 mm apart along the longitudinal axis and lofted to produce a tapered prismatic form factor inspired by traditional Wa-bocho knife handles.

Global

The resulting solid body was then hollowed using: Modify → Shell

Wall thickness was set to: 3 mm

This produced a uniform outer shell while creating an internal cavity that can later host an internal lattice structure.

The purpose of this internal cavity is not only to:

  • reduce material usage
  • reduce part weight
  • interrupt conductive heat paths between the dome and the user’s hand

This geometry cannot be manufactured subtractively due to lack of internal tool access.

Global

This approach enables the later addition of a fully enclosed internal honeycomb lattice that will:

  • be trapped inside a closed shell
  • have no machining access path
  • demonstrate additive-only manufacturability

This directly satisfies the Week 5 assignment requirement of designing an object that cannot be produced using subtractive methods.

Global

STL Export

Once shelling was completed:

  • The body was exported as an STL file
  • Prepared for slicing in FDM workflow

The internal lattice patterning will be implemented in a later iteration as part of further design refinement.

References

This week combined several sources of knowledge including Fab Academy lectures, local class instruction, student work, and experimentation with AI-assisted workflows.

Fab Academy Global Lecture - Adrian Bowyer — Introduction to 3D Printing and Additive Manufacturing

Fab Academy Local Instruction - Fab Lab Barcelona instructors and class demonstrations - Clay paste extrusion session - SLA printing and post-processing workflow

AI + Digital Fabrication Context - Cesar Garcia — RAGLab presentation (local retrieval pipelines for fabrication knowledge) - Amira Abdelrahman — AI & Design lecture (AI integration into constraint-driven design workflows)

Repositories - FabRAG / RAGLab repository
https://github.com/academany/FabRAG

Technical References - The 3D Printing Handbook: Technologies, Design and Applications — Ben Redwood, Filemon Schöffer, Brian Garret - Fab Academy weekly documentation examples - FDM infill and lattice optimization references

Use of AI

AI was used throughout the week primarily as a thinking and documentation tool, not as a source of final engineering decisions.

It supported:

  • translating fabrication constraints into CAD workflows
  • structuring documentation
  • generating modeling instructions
  • exploring additive-only geometries
  • testing local AI + RAG workflows for fabrication knowledge retrieval

Example Prompts Used

Additive Manufacturing Design

“Explain the key geometric constraints when designing parts for FDM printing, including overhang limits, anisotropy, and strategies to avoid internal support structures.”

CAD Workflow Prompt

“Generate a Rhino and Grasshopper workflow to model an octagonal Japanese wa-style handle designed for FDM printing with an internal lattice structure and a central rod bore.”

Fusion Modeling Assistance

“What is the best workflow to loft between two octagonal profiles in Fusion to create a tapered handle suitable for 3D printing?”

Lattice Design Exploration

“Compare honeycomb, gyroid, and triangular lattice structures for reducing heat conduction, lowering weight, and maintaining stiffness in a cylindrical handle.”

Fabrication-Aware CAD Prompt

“Design an additive-only internal structure inside a hollow handle that cannot be produced using subtractive manufacturing.”

Local RAG Experiment

“Generate fabrication-aware Rhino modeling instructions for a wa-style octagonal handle with an internal lattice based on Fab Academy Week 5 additive manufacturing constraints.”

This experimentation also helped reveal the limitations of local RAG pipelines when fabrication knowledge is not properly structured or embedded in the dataset.