Feasibility of using color database to identify manufacturer of unknown filament

ch_charlie

So, the filament library helps people find filament that matches the color they're looking for. My question is for the opposite use case: Is it possible to scan the color of a 3D printed object, look up its hex code in the library, and then extrapolate what brand/manufacturer made its filament by finding the matching hex code in the library?

This is more to bounce ideas than anything - I'm researching ways to recycle 3D printed plastics. A lot of makerspaces and clubs can't sort their own waste properly, so we need a way to identify unknown 3D printed objects in order to separate PLA parts from PETG parts, PLA from ABS, etc. Different polymers do not mix very well together and usually cause recycled filament rolls to snap or clog when printing.

It seems every filament has a unique hex code. So if I scan a printed object with a color of, say, #FE6E32, it's probably Prusament's orange PLA (#FE6E32). Nobody else make the exact same color, so I can be reasonably confident that the part I am holding is made of PLA. If I see a second object with a color of #EB5405 (Prusament - Prusa Orange PETG), I know to separate it from the PLA so they won't mix.

Just some food for thought.

jkaufeld

Personally, I don't think this method is worth pursuing. The quality of the measured color varies depending on how good your sensor / equipment is, and unfortunately, color does still change by batch. Not much, mind you, but it does change. The amount that it can change would be over the threshold of the difference between something like PETG and PLA.

Semi-translucent filaments especially would be absolutely awful to figure out via a color reader, since the difference between a tinted PETG and a modified transparent PLA is not far. Sorry, I wish I had better thoughts to share.

ch_charlie

jkaufeld Thanks for the input.

Most clubs I spoke to were willing to share their filament order history to help facilitate recycling - do you think I'd have a chance if I knew in advance of the color changes (either by sampling their incoming rolls, or by collaborating with the manufacturer to be forewarned of batch changes)?

Actually, if I know the exact filaments in use, that really narrows down the list of colors to check for a match. One club only carries 3 PLA rolls- black, red, and white from Elegoo. Anything in their bins will only be one of three colors. Surely that'd leave a decent margin for error.

Other clubs might be more difficult, like the local university lab which carries 36 different filaments. But obviously I would not mix the bin with those 36 different colors into the first club's bin of just 3 colors. At least some of them should be sortable.

For those translucent rolls and other problematic filaments, I guess I'd have to sort those out by hand.

ch_charlie

jkaufeld

Or if one could control and measure the filament color as it's being made, wouldn't it be trivial to identify your own filament from a pile of other plastics? I might be able to find a makerspace or lab with a Filabot machine or other plastic extrusion equipment nearby, I might give it a try just to test the concept if I get the chance.

jkaufeld

ch_charlie Mmm! If they're willing to share their order history, then yes, this has a much higher chance of working. I was imagining this from the perspective of "I get a box with scraps from an unknown person and an unknown materials list and have to figure it all out", not that you'd have a good idea of what's coming in. I think that you will run into issues identifying the difference reliably between PETG and PLA of the same official color, though; a controlled burn / melt test might be the only sure-fire way to tell them apart.

There is always variance in the measured colors, so you'd want to figure out ahead of time how much variance you're willing to accept for measuring what is supposed to be the same color. #6DCDE6 and #72CBE4 look different visually when you're comparing codes, but they're only a fraction apart in real life — how much variance is a good amount?

As long as you're working with specific groups and have that order history, then yeah, I feel like this would work and work decently. Let me know how it goes!

ch_charlie

jkaufeld Will keep you posted!

Yes, two different polymers that share one color would be an issue. For that scenario, I'll have to find other tests.

I did a bunch of research and now think infrared spectroscopy can help. Different polymers have different "fingerprints" in the infrared spectrum, and there are some open-source databases for that.

It's not perfect - IR can't tell apart wet or dark-colored plastics, for example - but it doesn't have to be. As long as it can pick out some plastics in the bins I collect, it's an improvement over sending them all to landfill.

It seems like I just need to find the right "clues". For example, with black-colored plastics that confuse both visual and IR sensors, there might be other clues like hardness (PLA is harder than PETG), density (some plastics float in salty water), magnets (if the plastic has metal additives in it), and so on. It doesn't have to be super accurate, just precise enough to say "this plastic doesn't match the others". And if a makerspace print farm only has a dozen or so plastics to check, that should be good enough.

ch_charlie

Oh, and there might be other clues that have nothing to do with the plastics, but the print user who sent them to me.

If I get a box of plastic from somebody who does not have an enclosure, that box probably won't have fuming plastics like ABS in it, unless they're crazy or unaware enough to risk plastic poisoning with those. However, if the same person does have a hardened nozzle, then I should probably check for abraisives like glitter (look for sparkles), carbon fiber (absorbs radio waves), and so on.

I'll let you know how it goes! I feel like I'm really close to finding something that might work.