Trying to dial in a large DIY heated printer

[coseng 4]

I love the embedded nut idea!  The multiple pauses is dedication!  Great idea and flag choices,  😉

 

I'm going to test the parts on track as-is.  There is a lot to be tested for rider ergonomics and maybe even some wind tunnel time.  Once I am happy with the entire design, I'll have an autobody shop (lots of cars use ABS bumpers, etc) sand and prime to get a great surface finish then give them to the CF guy who will make molds from them and make final CF parts.  Acetone smoothing works on small stuff but think these large, gently curved surfaces will end up looking like warped glass and I don't think will make the final finishing any easier.

 

Made and ran a PA test part with geometry similar to my problem areas. This part's toolpaths are one continuous pass per layer, no infill, retractions, etc. 0.8mm nozzle, 0.5mm layer height. 135mm/sec actual print speed.

It printed mostly great, with the exception of one area:

The location of this problem points strongly to thermal issues of the extrusion bead cooling, so next on deck is the layer fan.

The PA test changes were subtle and easier felt with your fingertips and not a camera. Best value seems to be 0.03 where there was the least amount of bulging at the rib ends but no other weirdness at corners and transitions.

Progress!

Edited June 4, 2022 by coseng

[coseng 4]

Oh yea! Finally after running this part about 20 times, here's a great final result. It is not the first white part, so was not a material swap fix.

It's just about as perfect a part as I can expect. The grooves on the backside are for 1/2" square steel tube for a small reinforcement frame that will be epoxied in and bolted inline with the two mounting holes so these parts will actually be used on the racetrack at up to 160mph.

 

The last few final changes were all thermal:

-added a layer fan, off for first 30% of build, then increasing to 75% at 75% of build.
-reduced the chamber temp as the build progressed, 50C to start, 45C midway, then 40C at 75%.
-increased minimum layer time from 10sec to 20 sec. Build time was 7hr2min.

 

My take away from all this testing is that although it is often said to use minimal layer cooling on ABS, I suspect that this is for machines without a good thermal enclosure, which seems to make a big difference. I think that my air/temp mixing in the chamber is not that good, or the thermal gradient is just too big to keep constant with a thin bellows top with a gap for the printhead, or my sensor positions are reading low, or all of the above. With a 50C chamber setpoint, the bed sensor reads is 93 (set for 90) and the SSR never goes on. With a 45C chamber the bed is 92, with a 40C chamber, 91. So I am getting great heat retention to the point of excess, where the chamber heaters are heating the bed. Which really just means I have to figure out those numbers for this printer. Or maybe the heater tuning should be with all configurations of the various heaters on and off, but then it gets pretty complex and more likely to mess things up.

 

Once things get up to temp they are quite stable so I don't want to mess with it too much. So I’m getting there. Hopefully a revised layer fan will reduce the amount of heat soak, for lack a better term, and allow me to do 10sec layers instead of 20sec. Or I’ll just print two parts at a time. But for now, it is 235/90/50 to start, then dropping to 230/90/45 at the midpoint, then 220/90/40 for the end. Max speed on the big layers is 150mm/s, down to about 40mm/sec on the small layers.

 

Thanks again to all for the help!

[GregValiant 1,246]

Congrats!!!

I had a feeling it was a thermal issue.  I wanted to back away from the theory so I wouldn't get tunnel vision but it was sticking in my head.

 

If you had told me it was going to go by me at 160 I'd have told you not to worry about how it looks.  🏍️

[coseng 4]

Thanks!  There's still more speed to come, but for now I need to print some parts and not bother about pushing the limit. 

 

Have a few good runs under my belt and am feeling pretty good about the settings. I revised the layer duct fan to blow a bit more and reduced the minimum layer time to 15 sec with no apparent change in the good results.

 

Still not quite sure I understand the thermal relationships going on. It takes setting the chamber temp to 35C for the bed heater to actually cycle. At 40C chamber or above it reads a couple of degrees over the setpoint of 90C, and continually reads 92 or 93.

 

Wondering if in the duet 6hc or rrf firmware if there some way to log or graph the activity of all the heater circuits to see what their duty cycles are.  Maybe the good insulation and almost 80W of heat input from the printhead is enough to keep the chamber up to temp, as unlikely as that seems. At 50C chamber temp the outside surfaces are warm to the touch and the glass window pretty warm. At 35C chamber temp the outside surfaces or door window do not even feel warm.

The step in the airbox part is because it was being run with no supports even with completely horizontal overhangs to see how bad the bridging was. A blob developed at one overhang and hit the nozzle and shifted the bed a little but it kept printing. I only wanted the first couple of inches anyway to check the assembly, since that is where everything mounts to.

The marks on the back of the fairing parts are where I trimmed the ribs off but did not do a full finishing job. I'm very happy with the surface finish of the appearance surfaces. They have a consistent satiny appearance.

Now that I am feeling confident, one half of the gas tank/airbox cover is printing. Multiple smooth curvatures, some near vertical, some near horizontal. So far, so good. Projected to be a 13hr print and about halfway through. This is the first part with supports. I used the normal style as the tree support option resulted in a 2hr longer print time. Maybe I didn't tweak the settings optimally.

If that goes well, next I'll do the tailsection standing up, which is about 780mm tall, and really see how good the Z alignment, straightness, parallelism, etc are. I could print a large spiral vase box or cylinder and see quickly but that would be too easy!

Chris
Cosentino Engineering

[GregValiant 1,246]

I think if the "squareness" of the printer is off it won't be by much or your previous parts would have shown an indication.

Regarding the temperatures - Are you confident in the sending units/thermistors calibration?  Is it possible they aren't just measuring the device they are supposed to monitor but picking up some residual heat from another source (like the bed isn't that hot but the thermistor is affected by another device?)

 

When you get a minute, step back and take an overall picture of the whole gizmo.  The biggest that's been posted here is this one I named Godzilla.  It uses a big ole pellet-fed extruder.  He was doing flat panels and wanted to print them in Vase Mode.  I helped a bit in redesigning the parts so it could be done.

 

 

[coseng 4]

>>I think if the "squareness" of the printer is off it won't be by much or your previous parts would have shown an indication.

 

Hopefully you are right!  Going from 300mm print height to 900mm so we'll find out.

 

>>Are you confident in the sending units/thermistors calibration?  Is it possible they aren't just measuring the device they are supposed to monitor but picking up some residual heat from another source

 

The temp sensor readings agree pretty closely with my handheld IR temp gun.  There is some cross talk, the chamber heater hot air exit has to be impinging on the bed somewhat.  The bed sensor is embedded in a 50mm long drilled hole from one edge, so not exposed to air.  I was just surprised at the seemingly small amount of heat needed to keep the unit at temperature.  Once past warmup, the nozzle heater is mostly on and the chamber heater is mostly off, and the bed heater is usually never on.

 

That's a nice unit!  What is your throughput with the pellet feeder?  I'm getting roughly 200g/hr with my wall print speeds of 150mm/sec.  The infill is usually too short to get up to those speeds so is about half that.

[GregValiant 1,246]

No, not mine.  It was the build of a poster here.  I tried to contact him afterwards but got no response.  The thread is HERE.

He stated in his post "...with pellet extruder, output around 2kg/h" so he's just a tad higher.  At a layer height of 8mm and a line width of 12mm he was going through a lot of material in a hurry.  I believe that works out to 4.62mm/sec (with a fire hose for a nozzle).  You would need to speed up to ~550mm/sec.  My Ender would need to run around 5,550mm/sec.

Things get out of perspective in a hurry.

[coseng 4]

That's some nozzle!  The clearpath tech was tuning the XY motors on my setup and they rapid smoothly at 1200mm/sec, but at those speeds the bellows is a noisy mess!  I limit it to 600mm/sec and it still seems like an instant move.

[coseng 4]

The tank finished successfully and is a mostly a really good print.

 

The only issues are on the horizontalish areas where I think another inner wall or skin layer is needed.  I'm not sure what the setting for this type of geometry is.

 

 

I was pleasantly surprised with how easily the standard supports broke off and the finish under them is still halfway decent.  I used the standard supports with I think too fine of a density.  Choosing the tree suppports (without changing any of the settings) increased print time from 11 to 13 hrs so I went back.  I think a bottom skin is also needed.  Those surfaces are completely internal so does not really matter but it is nice to have 100% nice parts.

 

I then went tall and tried the subframe, but with minimal supports, which ended up being a mistake.

One out of two legs printed fine, so this could still be used as a mockup and I am letting it finish. On hindsight, I was suspicious of the small islands with nothing joining them to the main brim but decided to run it anyway. I'll add some supports there and tie them to the main support structure before rerunning it.  Time to learn how to do that!

[coseng 4]

Also, just some little issues I'm tying to resolve:

 

The prints are coming out fine, but I am having one issue that requires me to babysit the first two layers. On the fist layer, the filament sometimes stops feeding. I notice it by the spool feed stepper not doing its chirp to feed. With my filament feed setup I can give a little manual push to the filament, then it starts feeding fine again. This happens 3-4 times on the first layer, then not again during the rest of the print. Each time it happens I have to blow off the pinch area because there are a lot of filament shards. Once the first layer is finished it does not happen again and as mentioned, at the end of the print the filament pinch area remains completely clean of debris.

 

The bed is flat so the nozzle is not getting blocked. I test it at temp using a plunge dial indicator clamped to the printhead carriage and get about 0.05mm total variance. I tried using the BL Touch for bed mesh, but was getting hugely varying scans that did not correspond to a mechanical check with an indicator. Now I just use the BLTouch for Z homing and it seems to preform repeatably and I print with no bed mesh compensation.

Also, FYI there is about 0.8mm difference in Z homing between hot and cold states.

[coseng 4]

On 6/4/2022 at 5:33 PM, GregValiant said:

>>The nuts go in during a pause as well.

 

One minor upgrade to this technique is to use a weld nut, like mcmaster.com 98697A810, since the larger rectangular size provides a lot more resistance to deforming the plastic and spinning.

 

[GregValiant 1,246]

Weld nuts, vibratory feeders, and lizard tongues (to place the nut at the weld location).  We used them a lot in auto assembly tooling.

I just don't need the size (or torque) of a M8 or M10 for what I do.  I use McM brass push nuts a lot but only just the #6, #8 and #10 sizes.  Just stick them on the end of my little soldering iron and in they go.  It's a piece-of-cake putting in the tapered holes too.  Honda (probably everybody else too) uses collared washers in the screw holes so the torque goes onto the collar rather than squishing the plastic under the screw head.  

Edited June 9, 2022 by GregValiant

[coseng 4]

These are M4, so like a #8.  I always end up with spun inserts.  These need a little more space but are a lot stronger.  And yes, those shouldered screws are used all over motorcycles for bodywork and airbox atachment.

 

The tailsection part was very satisfying to pull off the bed.  It came out mostly nice, but still some of these gaps in the skin issues.  I think the skin removal width and skin expand distance settings need adjustment.

 

 

The top had a bunch of oscillation residue from printing such a tall and slender object, but the part is still great.  Cura did a good job of slowing the print speeds down, but the accelerations and rapid moves were still high, which caused a visible wobble then the printhead did any rapid moves.  Maybe a quick script to reduce those accelerations and top speeds for tall parts.  I could also program a pause and glue in a temporary brace.  All of the surfaces are visible so I'd rather not add then have to cut away a big rib like I do on the side panels where one side is hidden.

[coseng 4]

Also doing a double run of the side fairing and was able to get some images of the first layer, where the infill looks a little sketchy, and may be clues to why there are some feed issues with that layer.    The problem does happen with some of the brim passes too.  For subsequent layers the extrusion bead look very clean.

 

[coseng 4]

Oops, had initial layer flow rate set to 120% for some reason.  Switching it to 100% and will be running another part overnight.

[GregValiant 1,246]

"...but still some of these gaps in the skin issues. "  I have a different take on that.

As the angle between the build plate and the part gets smaller, the step width increases.  I can't tell how many skins that area has (2?), but if it's bridging over air then there isn't enough squish because there isn't enough support for succeeding layers.  Maybe a support blocker configured as an Infill Mesh with top or bottom skins would add more material to the areas and allow them a better chance at printing correctly?

 

[coseng 4]

I think there is only one skin on these prints.

 

>>Maybe a support blocker configured as an Infill Mesh with top or bottom skins would add more material to the areas and allow them a better chance at printing correctly?

 

I guess I need to do some reading of the user manual!  I've been dropping in a part and slicing/supporting it with little to no manual intervention.  Admittedly, some of this printing is experimentation to see what I need to learn, and this is the first thing.

 

I printed the other half of the tank cover with reductions in the 'skin removal width' and 'skin expand distance' parameters after feedback on the duet forum and the results are definitely an improvement and I'll go some more with the reduction.

The overall print quality was largely the same (very good) but some of the areas with gaps closed up completely and the rest closed up a lot.

 

I also made the support structure a but more coarse and noticed a decrease in the underside surface quality at the benefit of about 2 hrs less print time.  Next time I think will be the fine supports.

 

So, definite progress, and only a little more to go!

[coseng 4]

Getting some really nice big prints out of the machine and even though I know there is still more slicer tuning to be done I just need the parts too much to slow down and do some careful tuning. Oh well, the many prices of success!

 

I now have the Cura 4.13 settings to 90% flow everywhere and 82% on the first layer and no filament tending is needed, at least for these two recent prints. On the first layer the brim top surface finish is quite nice, but the infill still is a little squishy at the ends where it merges into the perimeters. I'm not sure, but it seems to be overprinting some areas. I only notice it while tending the printer at the start of the cycle because these areas are covered by printing.

 

 

Tha areas that are least acceptable are mostly the bottom surfaces that are above supports and the bottom edges of some transitions that do not require supports.

 

I know that with a larger nozzle, big layer size, and fast print speeds there are limits to what can be achieved, but I am not sure how close to those limits these results currently are.

[GregValiant 1,246]

Always go slow when approaching the edge of a cliff.

 

The front end looks like it's off a Confederate/Combat Wraith.

[coseng 4]

>>The front end looks like it's off a Confederate/Combat Wraith.

 

I know this is meant as a compliment, but my front end is s a lot more advanced than the Confederate stuff.  And it actually works better than telescopic forks.  IMO he's a great artist, not so good a bike designer.

[GregValiant 1,246]

Well I guess I shouldn't expect too much for $155,000.00.

[coseng 4]

It's rolling art so can't be judged like a motorcycle.  But for that price you do get quite a lot of custom, beautifully machined parts and high end components with your piece of art, and it does ride, somewhat.  😉

[coseng 4]

OK, first big print failure. this was a sizeable part, 510x580x820mm that would have been about 5kg. I stopped at about 600z.

I paused it after about 300mm and was noticing some of the part sections swaying back and forth and tried to tape them to the more stable tree support sections, and that helped somewhat, but was not enough in the end. I need to add some more support in CAD that is attached to the part and will be cut off. It seems crucial where there are segments of the part that start on different islands on the bed then merge at some higher Z point. If there is any misalignment around the merge layers, the printhead hits one and knocks it out of place and then a cascading failure starts.

 

Before pausing to add the reinforcing tape I did notice the print nozzle hitting some of the part and tree support areas as it rapided over them. This makes me think that I do not have the feed amounts properly calibrated and am stacking a bit high on Z and causing the printhead to hit and knock it out of place. I can also add Z hopping but would like to get the physical characterization of the extrusion correct first instead of just adding band aids on top.

[GregValiant 1,246]

"...would like to get the physical characterization of the extrusion correct first instead of just adding band aids on top." 

Do you really want to start fooling around with the flow at this point?  I agree Z-hopping is a band-aid but sometimes all you need is a band-aid rather than major surgery.

 

Since a printer is a simple serial device and since I know a bit of gcode and a bit of programming I've written an application to control my printer from the computer rather than having to use the LCD knob (for movement, printing, whatever).  As the app evolved I added post-processing scripts to alter gcode files.  One of those post-processors alters the Z-hops in a file.

If I have a model that has a large horizontal hole, I know that when the hole starts to close up at the top "feather" edges will be created.  Those feather edges always curl and so Z-hopping is in order to prevent the nozzle from hitting the edges and knocking the print over.  I don't need them in the rest of the file, just for the twenty or so layers around the feather edge.  So post-process the original gcode file and remove all the Z-hops from this layer to that layer.  Then I run it again and leave the Z-hops in for 15 or 20 layers and remove them from the rest of the file.  (Since I am working independent of Cura I actually read a gcode file, alter some lines, and write everything into a new gcode file.)

An alternative is to create a gcode file with Z-hops and one without and then cut and paste them together (and to toot my own horn, I have a post-processer for that as well.)

 

When your small islands are laid down on top of the supports they may be like my feather edges, curling up as they cool.  I have my doubts that flow adjustments to fix that would have a positive effect.

For much of your project I think you are in uncharted territory but you see the whole project whereas from afar, I see snapshots once in a while and so my opinions may well be skewed.

 

One thing is for sure though...that's an impressive pile of spaghetti.

[coseng 4]

Yes, I will add the band-aid, but would like to know if I am still over-extruding or something else which is exacerbating the Z interference problem.

 

The printer is controlled using the Duet Web Control interface from my PC.  It is pretty full-featured as far as printer control goes.  I'm impressed with the ecosystem the Duet3d guys have created.

 

>>If I have a model that has a large horizontal hole, I know that when the hole starts to close up at the top "feather" edges will be created.  Those feather edges always curl and so Z-hopping is in order to prevent the nozzle from hitting the edges and knocking the print over.

 

Nice observation, thanks for pointing it out.

 

 

I've been CNC machining for a while, so also am acquainted with modifying g-code.  I've been inserting commands at varying z heights to control the heaters and layer fan.  Cura also has a nice post processor script functionality that can act as a second post to add modifications.

 

I have to do a little more thinking on how to do parts like this.  I think more CAD post work to add supports before importing to the slicer will be the way to go.  The geometry above some of these islands are not symmetrical and are off balance, which the slicer support generation does not account for.

 

So hoping a combination of z hop, better CAD support generation and maybe some extrusion optimization will resolve this mess.