Overview
Additive manufacturing, 3D printing specifically, has long since made its mark on the world. However, Selective Laser Melting (SLM) and Laser Powder Bed Fusion (LPBF) are technologies recently emergent. They are quickly becoming a more cost-effective and versatile method to print large, complex bodies unable to be manufactured by conventional means.
The quality of the resultant part is largely defined by laser parameters and the scan path algorithm used. I will be focusing on the scan path algorithms here, as that’s the main focus of my work.
If you’re interested in the research supporting everything below, you can get a download of my Mendeley library here. I’m pretty sure that doesn’t have PDFs, but it should be a good starting point regardless!
About Me
I’m a student at The Ohio State University doing software development for a place on campus on the side, where I focus on the development and implementation of new 3D printing algorithms for LPBF/SLM. At the time of writing this article, I am developing some submissions for the OASIS Challenge. After it finishes in March of 2021, you can check them out on my GitHub here.
My motivation for writing this article is to compile a wide variety of these scan path algorithms into one place, because they are called a billion different things and, when I was starting, it was really difficult to figure out what was worth it and what wasn’t. While I won’t act like I have all the answers, and I’ve only been working with them for a few months, I think I have an above-average amount of knowledge, and more than enough to give an overview of the common ones.
Line Scan / Meander
The most basic scan path algorithm; it is generally the default on most machines. You go back and forth across the entire part. Pretty simple, right?
From the literature I’ve read, this strategy has moderately good (not exceptional) performance all around, and is a good strategy especially for if you are just trying to get a bunch of material down. It is also dead simple to program, which is likely why it’s so common.
This is commonly augmented by reversing the direction of the hatches for ones that are side-by-side, or by doing multiple passes with skipping certain ones to lower the heat differential, as illustrated here:
From my reading, these are generally positive additions when it comes to print quality.
Paintbrush
You split your vectors along divider lines a certain width apart.
This can help a lot with minimizing the size of your vectors and bettering your part quality. I haven’t seen much research with this specifically. It likely borrows many of the same benefits as island scanning, discussed next.
Island / Chessboard / Checkerboard
An algorithm that splits the part into square sections, then does those hatches square by square, going among the squares in a roughly random order. This is basically paintbrush, but along two axes. It also seems much, much more common than paintbrush.
From literature I’ve read, this is generally a quality improvement over line scan.
Common modifications to this include:
- Offsetting the island grid system for each layer (1mm per layer seems common)
- Alternating the direction of each island layer by layer (e.g. if a layer’s hatches are horizontal on a given layer, they will be vertical on the next layer)
- Overlapping the edges of each island slightly, resulting in more heat input on the edges which can help smooth out the weirdness that can happen on island boundaries
Island scanning generally seems good in circumstances like line scan, where you are trying to get a bunch of material down quickly and there isn’t really any structurally delicate parts.
Spiral
An algorithm that spirals in from the outside.
While this algorithm can provide good performance, from my reading it commonly results in overheating the middle of the part, which can get pretty bad. Therefore it doesn’t seem like a great option.
This can also be done from the inside out, but from my reading that doesn’t really change any of the underlying heating issues.
You may also find contouring as a hatching technique. Normally contouring is just done on the border of the part, which helps with surface quality among other things, but you can keep inlaying it on itself. From what I understand, this performs similarly to spiral scanning, so I’ve included it here.
Fractal
An algorithm that borrows space-filling curves from mathematics and uses them to… well… fill space! This is super under-researched, and I’ve only been able to find a few papers out there.
Hilbert Curves (Left) and Peano-Gosper Curves (Right)
These boast very small vector lengths, which is generally a big positive insofar as quality is concerned. These also have a lot more mathematical basis behind them for why they work.
There’s too little research for me to conclusively say anything, but from my research these seem very promising and result in very good quality.
Looking For More?
I’ll be working with this stuff the next year or two, so feel free to ask me questions about anything (related to this or not) at andenacitelli@gmail.com, keep an eye on my GitHub (@aacitelli), or follow me on Medium for more to come.
This article was originally written on Medium, then transplanted here once I spun up a dedicated site for myself.