UM2 dimensional accuracy of prints

[jbeale 0]

I just ordered an Ultimaker 2, and I guess it will be a while before it arrives (I got two receipt emails; one said UM2 has 6-8 weeks lead time and the other said 8-10 weeks). Meanwhile:

My intended use is to make mechanical prototype parts in ABS. I have done this already with a Makerbot Replicator 1, so I have a feeling for what to expect, but I am hoping the new machine will do at least a little bit better.

What dimensional accuracy should I expect from parts? Say I print a rectangular bar with CAD dimensions 5 x 10 x 100 mm. How much different from 100.0 mm would I expect the actual length to be? If I print 10 of the same design on the same printer, how much variation might there be among the parts? Of course once I get the printer I can find the answer myself, but I'm just curious while I'm waiting.

I assume the dominant effect is the ABS shrinking and stress-relieving during cooling, which is dependent on geometry and % fill factor, resin composition, heater temperature, maybe even printhead speed (mm/sec) and other things. Are there any rules of thumb, or programs to calculate this? Or do users of consumer 3D printers generally not deal with tolerances?

Note: although a similar question appears in the FAQ

http://wiki.ultimaker.com/Frequently_asked_Questions#How_accurate_is_the_machine.3F

it only mentions the resolution of the printhead X-Y-Z movement. Resolution is not the same as accuracy, and printhead accuracy is not the same as finished part accuracy.

 

[conz 0]

What dimensional accuracy should I expect from parts? Say I print a rectangular bar with CAD dimensions 5 x 10 x 100 mm. How much different from 100.0 mm would I expect the actual length to be? If I print 10 of the same design on the same printer, how much variation might there be among the parts? Of course once I get the printer I can find the answer myself, but I'm just curious while I'm waiting.

 

Shrinkage of ABS is 0.4-0.7%. So your 100mm in CAD should be between 99.3mm and 99.6mm in your printed object.

It has nothing to do with the printer, it is a characteristic of the material.

EDIT: I have never printed with ABS.

EDIT2: Variation between parts should be very small.

 

[Dim3nsioneer 557]

First, congratulation on your ordering and welcome to the forum!

The dimensional accuracy is a very interesting question. As I do not print ABS I cannot give you proper numbers for the accuracy to be expected with this material. With PLA, which has a substantially smaller shrink effect while cooling down, I got positioning accuracies down to 0.1mm on nearly any dimension, coming mainly from the play of the belts. However, bores will always be up to half a millimeter smaller in diameter than they should and outer contours 0.1-0.2mm larger.

It's nearly impossible to predict. Or it would require a software similar to the software used for tool design in injection moulding. The usual way to go is try and error. Print the part once, measure the important dimensions and adjust the model, print again etc.

But this is the nice thing about printing with a consumer 3D printer, the costs of a print are much smaller than with a professional machine while quality differences shrink every day.

If everything is well adjusted and all screws are tight, then prints should be quite repeatable. Everything else is a hint that something is not as it should be.

 

[gr5 1,998]

Z dimensions are extremely accurate and should be within about .1mm. If you print in a heated chamber (recommended for ABS) then there will be shrinkage from the heated chamber temp to room temp.

XY dimensions will shrink for PLA about .3%. Not sure about ABS but I think it's about double. I believe both materials have a similar density/temperature graph and it's pretty linear in solid, glass, and liquid states - there arent major sudden expansions or contractions at any of those transitions. So mostly you only have to worry about glass temp to room temp shrinkage which is much more extreme for ABS because glass temp is much higher.

Anyway - this doesn't answer your question.

The mechanics of the printer itself should give you around .1mm accuracy and better than that repeatability.

The speed you print, the acceleration settings, cause over/under extrusion (repeatably) on corners such that you might get a bump out on a corner if you print "fast" because the bowden has some sprint to it and the filament keeps coming instead of slowing down on the corners.

The most extreme innacuracies will be for "bores" or vertical holes. There are 3 factors that shrink these things. I always add a minimum of .5mm to all diameters. Sometimes even 1mm. The 3 factors are shrinkage, number of lines inscribed in the circle (fewer lines = smaller hole e.g. 4 lines gives you a square hole) and the rubber band effect when printing the hole that pulls the lower layer inwards as the current layer is laid down or the upper layer may even be printed more inwards - think tight rubber band being glued down into a circle with the glue not quite dry. This effect seems to be worse for PLA because it has a lower glass temp.

In other words, PLA does most of it's shrinking while still liquid/glass. ABS does much more shrinking after it is a solid.

Repeatability - if you don't change any filament, temperatures, fan speeds, printing speeds - should be at least .1mm, probably half that. So you can put chamfers and other compensating factors onto your parts and get a non-conforming part to print so much better and still print very fast.

 

[jbeale 0]

Thanks to all of you for these helpful answers. With the other model printer, I did notice that bore diameters would shrink, while outside dimensions were usually pretty close. I don't need to print at high speeds, so I can optimize for that, although it occurs to me that software ought to be able to compensate for corner bump-out effects at least somewhat. At any rate, better than 0.1 mm repeatability sounds very good.

 

[gr5 1,998]

although it occurs to me that software ought to be able to compensate for corner bump-out effects at least somewhat

 

Yeah - that would be the firmware. To do it in Cura wouldn't make sense as you need to know characteristics of PLA at various temps and length of bowden and all kinds of stuff.

There's been some more talk recently of making the firmware compensate for the delay/pressure of plastic in the nozzle. I don't think anyone is working on it. There is an existing feature that does this in marlin but it's, well, not quite correct implementation. But it does help really crappy printers quite a bit. Not sure if it helps the UM2. Never turned on that feature myself.