Ideas on STRATEGY for optimizing stringing parameters?
[cross posting to MINI and Mk3S forum because it's not machine specific]
Never really had serious stringing issues that a simple temperature adjustment or my trusted heat gun couldn't take care of. Now I'm finding myself doing some production runs with 3D Solutech Chocolate Brown with too much stringing to accept.
What I'm interested is NOT just a list of parameters to modify. I know that. I'm also not interested in anecdotal suggestions of specific parameters. It all depends on your specific material.
What I'm interested in is ideas (ideally experimentally verified one -- yes, I used to be a scientist) to optimize a systematic exploration of parameter space.
What I am assuming here is that the quality of the print (aka least amount of stringing) is not driven primarily by one parameter (happy to stand corrected on that assumption) but that there will be local optima defined by specific combinations of parameters.
Let's simplify for the sake of argument and posit that only temperature, retraction length, and retraction speed were relevant (which I know is wrong). That alone would create a three dimensional search space: Vary temperature, vary retraction length, vary retraction speed. IF temperature were the main driver, I could simply find the temperature with the lowest amount of stringing, then play with the other two dimensions to fine tune the results. However, I don't have any evidence to believe that that's the case. So while say 210 degrees may give me the best results at a constant 3 mm retraction, it's not inconceivable that 200 degrees with 4 mm retraction would give me better results. Or 205 degrees, with 2 mm retraction, and a higher retraction speed.
Of course there are plenty of algorithms to search multi-dimensional spaces for local maxima or minima, but in the end we have to translate those into experimental setups. Using a fake multi-extruder setup in PrusaSlicer I can test some of these dimensions in one print, such as retraction length on the x axis and retraction speed on the y axis. Others such as temperature will require several prints. And of course it's not just three parameters, which one may probably explore exhaustively, but more 6 or 7.
So I'm interested in hearing what strategies folks are deploying to optimize print parameters (with a particular interest in stringing). Anything different than linearly testing each parameter -- and potentially overlooking a multi-dimensional local optimum?
And of course if anyone can give me reason to believe that one parameter alone is driving the system's behavior and everything else is just nuances, I'd die a happy person, too.
RE: Ideas on STRATEGY for optimizing stringing parameters?
[...] What I am assuming here is that the quality of the print (aka least amount of stringing) is not driven primarily by one parameter (happy to stand corrected on that assumption) but that there will be local optima defined by specific combinations of parameters.
I went way down this rabbit hole a couple of years ago. I was getting fine stringing that was driving me nuts. Determined to find the solution, I started doing repeated series of test prints altering parameters for:
- retraction and de-retraction length
- retract/de-retract filament speeds
- wiping
- z-lift or z-hop
- coasting (Cura)
In order to ensure I was making good comparisons, I wanted to eliminate my filament as a variable, so I did the following:
- ensured the filament (Filastruder PETG in this case) was properly dry.
- calculated the proper extrusion multiplier for the filament by averaging and entering the actual filament diameter into my slicer, and adjusting the filament extrusion multiplier for consistency.
Then I started doing my retraction tests. The results were surprising.
Starting at the bottom-left with 0 retraction and 0 z-lift, I had terrible stringing. Increasing retraction by 0.2mm as the prints moved left-to-right, front-to-back, stringing was minimal. After additional testing, I realized there was little difference between 0.1mm and 1.6mm retraction. There was barely any stringing. Then I thought about it.
All of those measures are workarounds for stringing. Stringing is usually caused by excessive filament either oozing out the nozzle or extruded onto the print. Eliminate the over extrusion and the stringing drops way off. If your slicer settings don't match the real-world parameters of your printer and filament, it's either going to over- or under extrude. Therefore, the biggest bang for the buck is to calibrate your slicer settings to match the physical characteristics of the physical printer and filament being used.
Since having this epiphany, I make a point of calculating the extrusion multiplier for each brand and type of filament I expect to work with repeatedly. A 5-10 minute process is sufficient to average the filament diameter and do a test print, create a filament profile, and that's it. If I see fine stringing afterwards, it's usually an indicator that the filament needs drying. I make a point of drying the filament before "fixing" anything in my slicer settings. I can consistently get good results printing PLA, PETG and other copolymers with 0.2mm retraction, 0.4mm z-lift, and wiping disabled. I don't use coasting in Cura.
Some filaments like flexible are going to be problematic no matter what you do, but for most everyday prints, this has worked well for me.
I've put notes on my testing procedure here, and details on the "multiple virtual extruder" approach for testing retraction settings here if you're interested.
RE: Ideas on STRATEGY for optimizing stringing parameters?
@bobstro
Your page on the multiple virtual extruder approach was what actually made me start thinking about this whole issue 😁 I realized your setup was sequential (using the nine points to check nine values of one parameter) and started to wonder what if I modified two parameters across the two axes. Which made me then wonder how to address the other dimensions of the problem...
RE: Ideas on STRATEGY for optimizing stringing parameters?
[...] Your page on the multiple virtual extruder approach was what actually made me start thinking about this whole issue 😁
I was expecting to do a lot more iterations to first determine best result for retraction with no z-lift, wipe, or other mitigations, then start fine-tuning those other settings. I was really surprised when I realized I didn't have to. Once I made my slicer settings match the real world, it made sense that results would be better. As long as I do a bit of retraction to reduce nozzle pressure, oozing is not an issue. Tell the slicer to extrude a precise amount and if it has the correct parameters, the printed result should be very close to the slicer preview.
This is also when I realized moisture absorption can be an issue with any filament. If I've been printing for a while, then start to observe stringing, drying the filament has fixed the problem.
I realized your setup was sequential (using the nine points to check nine values of one parameter) and started to wonder what if I modified two parameters across the two axes. Which made me then wonder how to address the other dimensions of the problem...
I stuck to changing one variable per-print, but you could certainly try 3 rows with 3 combinations of retraction and z-lift for example. With a bit of fidgeting, it's probably possible to fit more than a 3x3 pattern on the build plate. The problem is that those virtual extruders are really tedious to work with in the slicer.
De