Thank You, Prusa for Sharing Failure Statistics
Most people probably overlooked this, but I think it's worth giving Prusa props for sharing some data I haven't seen shared by any other 3D printer manufacturer. Specifically:
The testing process of the MMU2S took 2 months to finish, we used 20 printers and we printed 478 test blocks designed to test frequent switching of all five filaments. There were two types of blocks: the first one took 19 hours to print, it has 607 filament changes and the second one took around 70 hours to print, while it had 3520 filament changes, so the total value of filament changes was nearly 600.000. Both blocks were designed to test frequent filament changes (even in higher layers). More than 93 % of this prints were successfully completed (about 19 % of prints required user attention, but they did not fail) and we are still working on the firmware improvements to decrease the number of failed prints even more.
I think every 3D printer manufacturer should share similar information about their testing process and failure statistics. Period, end of story.
I suspect we don't get this information because very few manufacturers do any kind of rigorous testing at all. I suspect Prusa didn't do it with any prior MMU iterations either. But testing and being open about the test process and results is an important sign of a maturing company that's trying to do right by their customers.
Aside: I can quibble with Prusa's test process. For example, they should have tested more than 20 printers, and they should have tested MMUs assembled by typical customers and not just Prusa engineers. We will soon find out if they overlooked anything important. But the important thing is that they tested and shared their results.
Re: Thank You, Prusa for Sharing Failure Statistics
So let's see. Someone who is good at statistics do the math for us.
93% success rate. 7% failure.
19% user action.
600 and 3500 filament changes.
So on well calibrated machine, we are looking at roughly a 75% success rate, start to finish. I guess that's not too bad.
The reason I don't include "user action," is because the print might continue and succeed, but it might have skipped a layer and been ruined some how... even though it "finished"
I bet my MK3 is probably somewhere in the 95% success range. I'd be willing to say 1/20 prints fail due to some random problem... bed adhesion, or shitty filament, jam, burn marks on an outer surface, or wrong Slic3r settings, etc.
Hmm... I guess its hard to read into the stats Prusa provided. But yes, kudos to them for being upfront about the success rate!
Re: Thank You, Prusa for Sharing Failure Statistics
So on well calibrated machine, we are looking at roughly a 75% success rate, start to finish. I guess that's not too bad.
The reason I don't include "user action," is because the print might continue and succeed, but it might have skipped a layer and been ruined some how... even though it "finished"
Prusa is saying: "93 % of this prints were successfully completed" . I would assume that missing layer doesn't count as successfully completed print. Otherwise you could question whole 93%. Missing layer are also happening without user intervention btw.
Based on my experience you can fix the "user interaction" cases without any visible mark on the printed object. Some do. For example power panic function leaves you with at least slightly visible mark and you need to press a button to continue.
I'm reading 93% is a real success rate and 74% does even work as "fire and forget". Still might be disappointing for some people if they are one of the 26% either failing or babysitting the print. But this statistic might encourage them to look into the root cause.
Often linked posts:
Going small with MMU2
Real Multi Material
My prints on Instagram
Re: Thank You, Prusa for Sharing Failure Statistics
well, if they have been using the extruder "improved" filament sensor I can see how that alone would "could" deal with allot of skipped layers.
I have seen skipped layers happen I know it wont be 100% but the 2nd filament sensor active should save some skipped layer issues. at least stop the printer from printing nothing. user interaction isn't a big deal. I would rather come home to a printer sitting there like a dummy, then come home to skipped layers.
at least if the printer stops the user has the option to investigate what happened.
I have a Prusa,therefore I research.
Re: Thank You, Prusa for Sharing Failure Statistics
I applaud Prusa3d for the MMU2s. I received mine, got a missing part quickly, and Sunday I put it into action.
I need to work through how I am not adequately set up. My failure rate for any particular filament change is about 5%. The test object I'm printing is an object from hell, with about 625 changes. Because the object is small, I'm averaging about 2.5 user intervention events per hour.
These are not focused on one path. They are spread among all 5 paths. They include both unload and load errors. So far, I can recover by following the user interventions.
Major KUDOS to Prusa3D for the firmware. It works like well-tuned, n-th generation code. It includes some automatic error recovery states. It recovers with user intervention without losing position or causing disruptions in the print.
I haven't found a significant difference in failure between Prusament PLA, Prusa PLA, Hatchbox PLA, and one other non-brand.
The end-nibs looked a little better when I lowered the temperature to 200C from the default 205C.
I will be watching the forums closely for ways to improve my error rate. As it is now, it isn't really useful. The Prusa3D i3m3 was a toaster. Everything came out just as I expected. This is very finicky, and requires that I be present, or have the patience for letting 20 hour prints proceed only during the times I can be there to clear errors.
Where should I look for advice in improving the error rate to be close to what Prusa test has revealed?
-- Carl