Laser Filament Monitor - test results

tinkerz · 8 Sept 2018, 22:44

@wilriker disregard, wrote a quick gcode file and figured it out, comes out on the WebUI. Perfect. Adding to end scripts. Now we just need logging.

Tinkerz

wilriker · 16 Sept 2018, 17:59

I switched housing again to the cone-shaped interior one on Thursday night and printed something on Friday. At the end of the print the statistics looked awfully low and I was already blaming the housing. When I inspected the prints, though, it turned out the sensor was right. I had massive under-extrusions especially on one layer because my filament spool had fallen off the rolling spool holder and the friction was too high for the extruder to pull where lots of filament where required quickly.

So, now I decided to enable the sensor for the current filament though I am still testing. Let's see how often my prints will be paused.

P.S.: A third option for the sensor would be nice. Currently there is just

ignore issues (but collect statistics) and
pause the print on issue detected

I would like to see a third option where issues would just send a warning but the print still continues.

KeeganB · 20 Sept 2018, 18:28

Hi All!

First, sorry it has taken me so long to join in on this thread! Life has been keeping me away from printing lately.

I am testing the monitor on a SeeMeCNC Artemis. As such, I designed a mount for it that attaches it directly to the EZRstruder. Here are a few pictures of the mount. I have only done one print with it, and everything seemed to work fine. I will get a spreadsheet going with my results. I will also upload the mount STLs to Thingiverse and drop a link here.

One thing that I realized after my first print, was that my mount doesn't have an enclosed side on the opposite side of the filament from the laser. On David's Thingy file it has a small extruded box there, but I didn't even think about that when modeling.

Well, there are photos there I think. I can't figure out how to do them I guess?!

Thingiverse Link: https://www.thingiverse.com/thing:3111798

brunofporto · 20 Sept 2018, 20:06

@t3p3tony said in Laser Filament Monitor - test results:

@guyaros That looks great. actually during test on the filament we used etc we got the best readings at 9mm gap but that is not conclusive. worth trying a variety of gaps.

I'll print the first with 9mm then

brunofporto · 24 Sept 2018, 14:20

I am designing my casing. Please, what is the recommended aperture angle for the sensor window/channel?

I remember reading it somewhere at this category but could not find it anymore.... Thanks!

brunofporto · 27 Sept 2018, 12:58

Here are my results so far.

The first two prints where made with Blue and Dark Purple ABS. One of the things that made me extremely happy was that these first two prints where slightly under extruded (little gaps at top layers). And as the average readings where consistent I used them as reference to tune the steps/mm. Based on the average (93%) I increased the steps/mm of the extruder from 85 to 91,4 and the next print was flawless.... With average of 100% and perfect top layers (IMO).

Then I started to test different materials and colors. The two translucent tests where with PETG and PLA. The curious thing is that besides the readings differ a lot from the reference blue ABS they ratio of predicted length to read length is better!!! While the ABS was between 94% and 98% the transparent generic PLA was better than 99%!

@T3P3Tony and @dc42 I am compiling all my prints at https://docs.google.com/spreadsheets/d/14YdQkZd0Ojso2ye2EpN1kS2Y_nZz_KdEbPJ7vjNBh50/edit?usp=sharing

I am planning to swap the housing to the reference one and then compare results for the same materials.

As soon as I get at least 5 prints per material I'll switch the sensor to test X and Y movement instead (not sure how to to that yet....).

Please let me know If you would like any specific test to be done.

fma · 27 Sept 2018, 13:14

Is you sensor before or after the extruder? In Bowden config, the marks on the filament seem to help reading the movement.

As I'm using a direct-drive extruder, I'm wondering if I could not build a system which draw some marks on the filament before the sensor, and clean them up after... Any idea what I could use as 'paint'?

brunofporto · 27 Sept 2018, 13:26

@fma I proposed that too, somewhere here I think

It was not recommended due to the possibility of fragments dirtying the sensor. Also it could buckle there... Then I decided to increase the exposed length of filament with the housing design.

But the readings are not bad! They are within a margin and the average is very consistent for each filament so far.

They could just design the decision logic based on the average variation... If it keeps within the average for some length it is ok... If the average starts to diverge too much than trigger the alarm

I only have 7 prints registered so far so... it may be too early to any assumptions. But I am really impressed by solving my under extrusion issue in one shot based on the average reading!!!

fma · 27 Sept 2018, 13:48

@brunofporto said in Laser Filament Monitor - test results:

It was not recommended due to the possibility of fragments dirtying the sensor.

Mmm could it be possible to find something which does not make dirt? Like a marker? Ideally, something translucent, but reacting with the laser IR, so no need to clean it up...

brunofporto · 27 Sept 2018, 13:52

@fma A marker would tint the filament.... It is even used for that purpose Like this: https://www.thingiverse.com/thing:11742

But it had no problems so far reading transparent or translucent filaments (one test each....) Besides % readings been different from the base blue ABS they even had a much better approximation of the predicted filament length by the slicing software.

fma · 27 Sept 2018, 13:58

Yes, that's why I suggested to use something transparent, but reacting to IR...

brunofporto · 1 Oct 2018, 07:53

@fma Well.... That should work but then there is the issue of finding a paint or marker that does not add color, mess with filament properties or is chemically innert to most of them, is opaque to IR and worldwide available

One alternative to that would be use the old hob wheel against a bearing and then read the bearing surface as this sensor is very good at reading stainless steel. Nut we would be back to the limitations, higher cost and issues of a hardware dependent filament monitor.

Another thing would be use some optics focus the sample filament length to the full sensor area available... But: expensive... complicated... can scratch and get dirty..etc....

I believe that with enough real world data and a good sample area they can tune it and suggest good margins.

kraegar · 27 Sept 2018, 18:05

So, I've tested printing a sensor housing, and then painting it in black 2.0 paint. An IR cam & IR light shows the black 2.0 is very IR absorbant. However, I've seen zero difference between the painted housing and the Carbon Fiber PETG housing I started with. (Which also seems to absorb IR very well on camera at least).

I keep forgetting to record the results after prints.

https://docs.google.com/spreadsheets/d/1a5suKtijUsatl3DhUFiox_rfW39qo9kHSlihYHaAcxo/edit?usp=sharing

fma · 1 Oct 2018, 07:53

@brunofporto said in Laser Filament Monitor - test results:

@fma Well.... That should work but then there is the issue of finding a paint or marker that does not add color, mess with filament properties or is chemically innert to most of them, is opaque to IR and worldwide available

I would like to make some tests, with different stuff. My idea is to use some sort of gear to put something at regular steps. What length of filament does sensor see? Just to design the gear pitch...

dc42 · 1 Oct 2018, 08:51

@fma said in Laser Filament Monitor - test results:

I would like to make some tests, with different stuff. My idea is to use some sort of gear to put something at regular steps. What length of filament does sensor see? Just to design the gear pitch...

My guess is not very much, because the laser spot should be small. Unfortunately the data sheet doesn't specify the beam divergence. It does say that the distance between the sensor and the target should be at least 3mm to get maximum resolution.

fma · 1 Oct 2018, 09:31

Ok, thanks. I'll try to make a small pitch...

Snypes · 9 Oct 2018, 11:49

This post is deleted!

Snypes · 9 Oct 2018, 11:53

Has anyone considered uncoupling the filament from the laser and having the laser read a constant metallic surface for example (much like an idler bearing maybe?)?

That's pretty much the solution used by jtx on his skelestruder and it would at least allow all filaments to be calibrated against a constant....

https://www.thingiverse.com/thing:3091625

kraegar · 9 Oct 2018, 14:40

I may have a new "least accurate filament" - Atomic Marble PLA

Duet3D laser filament monitor on endstop input 3, disabled, allowed movement 40% to 120%, check every 3.0mm, current position -1.1, brightness 96, shutter 17, measured minimum 12%, average 71%, maximum 167% over 92403.1mm

brunofporto · 9 Oct 2018, 15:32

I do also compare the slicer predicted filament usage with the measures filament usage and have a very unusual inverse relation where the ones with the worst reading % also have the most accurate measurement of total filament usage!