Printing TPU on the Bambu Lab H2D

A friend had need for a replacement drive belt. A good excuse to try out printing with flexible TPU.

The original belt plus the printed TPU belt.

I managed to pick the wettest day in months although luckily the humidity where my printers are only got up to 57%.  I was not expecting great results. I have not attempted to dry the filament. It's new, straight out of the vacuum sealed bag. I have an AMS HT on pre-order specifically for this purpose but that is not here yet and I still wanted to have a go.

I've used Bambu Lab TPU 90A filament through a 0.8mm nozzle, which I will dedicate to only use TPU. They are printed with a 100% rectilinear infill, solid. I've created the model of the belt in FreeCAD and deliberately put a flat base with 45 degree tangential slopes leading into the otherwise 4mm diameter cross section belt. I've sliced this with 99 wall loops so the result is solid and random seam locations for added strength.

There are contradicting instructions on Bambu's site. The filament purchase page in the store saying use glue stick for TPU and the wiki saying don't use glue stick on a textured PEI plate because it might stick too much. 

https://wiki.bambulab.com/en/h2/h2d-tpu-printing-guide

I opted not to use glue stick and I was easily able to remove the printed belt using an old plastic membership card, sharpened at one end. (I've since found that was not necessary.)

Even with a simplified filament path, approximately as instructed by Bambu but without a suitable container, the pull on the new heavy filament spool was greater than I thought suitable. I made a make shift bush out of a pipe but that barely helped. I thought about pulling out the length of filament needed and cutting, so there would be no spool to pull, but my solution for today was that I kept returning to the print while it was in progress to pull lengths of filament off of the spool by hand, so the extruder had minimal resistance.

The results, although not pretty, are very serviceable for the purpose.

This can only get better when more suitable kit arrives.

==

Update:

One of the belts has been put to good use and was left running the machine non-stop for over 2 weeks without an issue. The belt lasted another session before breaking.

Drying:

The AMS-HT arrived shortly after I had done the initial prints. I've now repeated exactly the same prints with thoroughly dried TPU 90A. The results are now perfect. Drying is well worth it with TPU.

I forgot to change the nozzle, so these belts were printed with a 0.4mm nozzle. Luckily, that's OK for TPU 90A. The results are very good.

They came off cleanly from the textured PEI print bed with minimal effort.

==

Blender Add-ons and tips for 3D printing

These are tools and techniques I'm trying out to help create meshes suitable for 3D printing. That is manifold and non-intersecting.

3D Print Toolbox

This is a useful set of tools for analysing a 3D model. It can identify if the mesh is non-manifold and  if it has intersecting faces.

https://extensions.blender.org/add-ons/print3d-toolbox/

Press the "Solid" or "Intersections" buttons to generate the statistics on the selected object.

I do not use the "Clean Up" tools. They are not as effective, in my opinion, as other methods.

Creating a Manifold Mesh

The most reliable solution I have found to create a single manifold mesh for 3D printing, is to use the Remesh Modifier. Sadly, that dos not retain the quality of the original mesh. It's perfectly OK for 3D printing quality down to 50 or even 30 microns (0.03mm.)

I often read about using a Boolean with self intersection with a cube. For me, that does not produce a manifold mesh when the original is made of multiple parts.

I have also tried Autodesk MeshMixer. In my opinion, Blender's Remesh produces a better result.

Joining Meshes into One Object (simple and quick)

This is the fairly basic method of adding several meshes into one object.

In Object mode, select the first object then use [SHIFT][LMB] to select others.

When they are all selected, press [CTRL][J].

They will all merge into the first of the selected objects.

Remesh Modifier

I have found this a fairly reliable method to make a tidy manifold mesh out of a mess. The result is suitable for 3D printing, however, none of the options retained the originals quality.

Remesh Modifier - Blender 4.3 Manual

In my experience, only a Voxel remesh creates a manifold non-intersecting mesh. The other options, such as Sharp with an Octree depth of 10 or above, produced a better quality result, but did not help for 3D printing.

Sharp with Octree Depth = 10

The quality depends a lot on the original model.

My preferred solution to create a manifold and non-intersecting mesh, suitable for 3D printing is to use the Voxel setting, with a size of 0.03mm. That is the thinnest layer that my resin 3D printer will produce.

Remesh Voxel Size = 0.03mm

Original Version

However, it's important to keep the original model, should I need to make any edits. The remesh is always the last process just as a means of satisfying the slicer software.

This gave good enough results, and the time to process was acceptable.

For larger, or complex models, this produced a lot of vertices. 34 million in one case. As I only usually print at a resolution of 50µm (micrometres), 0.05mm, for those models, I remesh at 0.05mm.

I'll continue to experiment with this.

For scenery, where the detail is not so critical, a Voxel Remesh, with a size of 0.05mm or even 0.1mm was sufficient and quick.

Apply the modifier to be able to check the result.

The Voxel remesh is also useful to make models used for boolean difference cuts. I have found that large faces often do not cut or join very well with the boolean tools. Remeshing can help.

Separating Parts of a Mesh

In Edit mode, select the vertices to include in another object of their own and press [P].

If the vertices are all connected, using [L] to select all connected vertices is a quick way to remove whole areas.

An Alternative to Individual Boolean Unions

Self Intersecting Boolean Modifier

This method leaves a hollow shell using a Cube with a Boolean Intersect modifier.

The result is similar to using boolean unions on individual mesh objects. In my attempts to use boolean unions on lots of detailed meshes, I found that it was very time consuming and would often fail. By using the cube with an intersection, all of the individual parts are made one in a single boolean operation.

Using the cube method does not create a manifold mesh, which is disappointing.

The method to create a single mesh is:

Create or import a model made of intersecting parts.

Join all objects in to one object using [Ctrl][J].

Create a cube larger than the object. Make sure it encompasses the whole object.

Add a boolean modifier to the Cube.

Set to "intersect".

Add the multi-part object to it.

And wait.

When done, Apply that modifier.

And wait again.

The result should be a single mesh but, probably, with lots of self intersecting faces.

Auto-Merge Vertices

This is off by default, but when trying to create or keep a mesh manifold, my prior experience is that it is easier if this is on.

In Edit Mode, top right of the window.

It does sometimes make unwanted connections but they are less than the confusion of the mesh looking visually correct but still having a hole in it.

It does this as you go, it does not fix all existing doubled up vertices.

Merge all Duplicate Vertices in an Existing Mesh

This is useful for fixing exiting meshes.

In edit mode, select all the vertices [A].

Use the right click menu or [M] to merge vertices.

"By Distance" is probably the most appropriate to use.

I work in millimetres so the default distance of 0.0001 is probably appropriate.

It's useful to have the Scene Statistics enabled to see it do something.

Show the Vertices Count, Scene Statistics

This is useful when carrying out several remedial tasks.

Simply [RIGHT-CLICK] on the status bar at the bottom of the Blender Window.

Tick to enable "Scene Statistics"

==

Mini filament review

For the last couple of days I have been printing some signs.
They are mainly flat with raised lettering.

Due to various failures I have been trying different types of filament among other things to fix the issues.

The results have led to some conclusions and preferences.

Filament choices in order of preference:

Rigid.ink PLA Plus
This is the best at the job so far. It stays flat on the glass with no warping even on the largest size flat prints.
It's not perfect though. The end of the spool is wound so tightly that the Ultimaker 2+ extruder struggles to get the last half dozen layers of coils off of the spool. That's a lot of wasted filament!
My guess is that the tight loops press against the inside of the Bowden tube so I can't even use it loose from the spool.
Heated to 225C with the bed at 45C.
The results have a nice tidy surface finish with sharp edges.
Takes spray paint well.
Update: Rigid.ink is no longer available. They suggest trying https://shop.3dfilaprint.com. I have used them before but have not got any in stock to try out. I'll update this article when I have tested some.
I'll probably try their Premium PLA:
https://shop.3dfilaprint.com/filaprint-black-premium-pla-285mm-3d-printer-filament-1380-p.asp


ColorFabb PLA/PHA
This was a lot better than standard PLA. It stayed flat on most prints with only a slight tendency to curl.
I get best result running it hotter than the suggested.

For me I heat to 215C with the bed at 60C.
The surfaces are nice and tidy. The raised details have a very slight bulge at the corners. If I was not comparing to other prints I would not have noticed.
When spray painting it is obvious the surface is porous along the print lines.

3DFilaPrint.com PLA
This would not stay flat. Even before it had done the third layer the edges of the brim were curling up and by the time the print had finished, the main body had lifted by a couple of millimetres at several corners.
I've used this successfully on other prints. It just does not like the thin flat prints.
Printed at 210C with the bed at 60C

Inconsistent Information
Several of the manufacturers provide guides or printed temperature ranges on their boxes. This information often changes. One manufacturer has three different heated bed temperature suggestions for the same filament, one on their web site, a different figure on a card supplied in the box and yet another range on the side of the box!

I no longer take the instructions as fixed rules. I use the suggestions as a guide to get me in the right direction. Trial and error is the only sure way.

==

Glass Thickness

As far as fixing the various issues, I think the glass is the main problem.

One of the problems is that the prints would warp. This always started in the rear right corner. For many prints that was the only issue but invariably the print would lift so high it would impede the print head and ruin the print.

I can only assume there is a slight inconsistency in the glass thickness. If I avoid printing on the rear right quadrant of the print bed, I get much more reliable results.
The affected area extends about 50mm from the right hand edge to about half way forwards from the back edge.

I managed to avoid this by rotating the largest prints and they have all worked since.

I plan to experiment with bed levelling to see if by deliberately setting a greater distance at the front left corner I can get better reliability at the rear right.

==

Worn Nozzle

Before I was able to get anything working I had to replace the nozzle.
The filament would flow for a while then get straggly. Small blobs every few millimetres with thin thread like strands between!

I don't know if the hot end nozzle was contaminated or worn but after I replaced it the flow was even again.

==

Printed with Thicker Layers

In the past I have found that thicker layers are troublesome. Now that I have done some more trials, I am able to successfully and reliably print with the following settings:
Nozzle: 0.4mm
Layer height: 0.24mm
Edge, top and bottom thickness: 1.2mm
Print speed: 60mm/s
Travel speed: 120mm/s
Adhesion: 7mm brim
Bed: Glass with hairspray (once in a while)

Rules of Thumb
I found on an Ultimaker forum, a rough guide to layer thickness and print speeds.
This approximately lines up with my experience, so I will use it from now on, rather than my usual guess work.

Maximum layer thickness: 3/4 of the nozzle diameter
e.g. 0.4mm nozzle can print a layer up to 0.3mm thick.

Maximum speed
The maximum print speed is determined by how fast the filament extruder can feed.
Stock Ultimaker 2 extruder, maximum throughput of 2.85mm filament: 7mm³/s

Calculation: Nozzle diameter x layer height x speed = throughput area
e.g. 0.4mm x 0.3mm x 40mm/s = 4.8mm³/s
Maximum speed calculation = Maximum throughput / (nozzle diameter x layer height)
Therefore the maximum speed is likely to be:
7mm³ / (0.4mm x 0.3mm) = 58mm/s

I have an Ultimaker 2+. According to the Ultimaker web site, the Ultimaker 2+, can print up to 16mm³/s with a 0.4mm diameter nozzle. It has a maximum travel speed of 300mm/s.
In theory, therefore, I could double the above speeds. In practice I have not been able to a achieve speeds much faster than the above calculations.
I tried, just as a double check. 0.4mm nozzle, 0.24mm layer and 90mm/s at 210C with PLA = 8.6mm³/s. It came out lumpy and did not stick to the bed properly.
The kit was clearly capable of pushing it at that speed but it would need a lot more tinkering to get close to something usable.


Nozzle Temperature
The hot end temperature is far more trial and error. That is partly because the thermostats in the printers are not that accurate.
Most manufacturers recommend starting at the low end of their suggested range of temperatures and work up.
According to most vendors, 2.85mm diameter filament tends to need to be about 5C hotter than 1.75mm filament for the same speed, so for a standard Ultimaker, which uses 2.85mm filament, start at 5C above the manufacturers suggested lowest temperature.
Increase the temperature for faster throughput. Alternately, it would be simpler to increase the speed for lower layer heights so the throughput is always roughly the same area per second, therefore the temperature could remain the same for all prints.

==

Warping and adhesion issues

Ever since upgrading the 3D printer to the Ultimaker 2+ I've had issues with prints warping and detaching from the build plate.

Today I finally twigged what was different between before and after the upgrade. The 2+ kit came with a calibration card for bed levelling.

I've measured the thickness of the card against the thickness of the paper I used to use. The paper is 0.03mm thinner at 0.12mm thick. Not only that but the new coated card has also picked up a thin layer of plastic in places from the nozzle so the thickness is no longer consistent, at one point is was 0.20mm thick compared to the average 0.15mm. The paper, despite having been in use for two years, does not have anything attached to its surface.

I've gone back to my thinner paper calibration and the first print is still attached to the bed.

I do not use any adhesive on my glass. I keep it clean with an imitation chamois leather and very occasionally degrease it with isopropyl alcohol.

I also made a change to the design of the model to add expansion gaps so that the shrinking plastic cannot pull the full width of the model.

Filament Guide Mark 6

I've now made several static guides for the back of the 3D printer.

The latest one works but I'm still nervous that the filament will stick causing under extrusion. To minimise that chance I've designed a guide that rotates using skate bearings.

I'm not claiming this is a new idea. There are loads of such things on both Thingiverse and YouMagine. None of those were quite how I wanted them or were fixed in a way I didn't want to do. I'm therefore working on my own.

This will hook under the lip at the bottom of the printer next to the power adaptor. There is no need to utilise any of the existing screws or holes on the Ultimaker.

Printed with supports for the slight lip

The bearings will hold the two halves together

Two halves interlocked

Printed in PETG

My aim is to minimise the angle at which the filament enters the feeder. I also want very low friction and minimal chance that the filament will slip out of the guide.

Bill of materials:
1x M4x40 bolt
1x M4 nylon locknut
2x M4x10 (Form C) washers
1x M4x12 repair washer
1x M8 washer behind the inner bearing over the plastic spindle
2x 8mm axle 22mm x7mm bearings (from skates)
Optional:
1x M8x1.6mm spacer between the bearings
3D printed parts:
2x pulley halves (with support from the bed)
1x Stand (no supports needed)
1x Support (no supports needed)
1x Spindle (no supports needed)

==

Downloads:
Filament Guide STL files (zip)


==

Links:
I have also made a spindle for the filament spool.

Nail sizes

Here's an odd post.

I'm getting fed up with large 3D prints failing after several hours of printing. I've decided to split some of them in to bits. They will need joining together and in some cases will need to be fairly strong.

I decided round wire nails would make ideal pins. The normal type you hit in to timber with a hammer.

They cost next to nothing each, when bought in bulk. There are lots of sizes. They have more than enough strength for my needs. I can cut them to length, if necessary.

I need to model a hole for them to fit in to, so I need the sizes.
These are the most common ones I have to hand in my shed.

6" = dia 6.3mm x 151mm, head dia 12.4mm x 3mm
4" = dia 4.7mm x 102mm, head dia 10mm x2.8mm
3" = dia 3.9mm x 77mm, head dia 8.6mm x 2.5mm (mild steel and galvanised were the same.)

Now I just need to use these in my 3D models.

Download:

3D model nails (blender)

UM2+ Under Extrusion

I've been having a lot of problems with under extrusion on larger prints. After inches of height I'll get some bands of poorly adhered fine stands before it goes back to normal.

I've been tinkering with temperature and feeder speed percentages and nothing is properly solving the problem. I had been printing some simple models in PLA at 60mm/s but perhaps it's over optimistic for that to work on every print. Slowing it down to 50mm/s has helped but I'd like to try to keep the speed up by making other adjustments, if I can.

Today I was printing in white and I noticed how many evenly spaced partial grinding marks there are on the filament.
My best guess from what I have read on the Internet is that these are likely to be caused by retraction being too frequent.

Filament Feeder Tension

As I'm mid print it is tricky to adjust retraction settings so I'm going to fiddle with the feeder tension.
This is on an Ultimaker 2+ so they come pre-set and are not supposed to need adjustment from the centre.

- One turn 

As it is only minor grinding, I've twisted the adjustment screw by one full turn clockwise. That is less pressure on the filament. The indicator moves up a tiny fraction away from the centre.

- Another turn

That's better but I still think the marks on the filament are too deep.
I'll try another turn.

From what I've read on forums and on Ultimaker's web site, if the filament is misshapen too much by the feeder it can either get stuck in the Bowden tube or in the hot end. Probably gets stuck at the isolator.

- One more turn for luck

I'm now three full clockwise turns off of centre and the marks on the filament are cleaners, showing just the even knurling and none of the grinding.

Reluctantly I've also reduced the speed of the print and at last I appear to be getting better layer adhesion.

I'm printing at 26% infill so I don't think that helps. At 30mm/s it's now feeding better.

==

Update: 1 August 2017

I spoke too soon. I was still getting under-extrusion just not for the same reason.

I'm fairly sure that the friction as the filament gets closer to the centre of the reel is still the cause of intermittent poor layer quality. I can either go back to using only cut lengths of filament, which is what I've done for years or I can use bearings for the the guides, axles or stands, as others do.

I'll have a look at what's available.

...

I fixed this by using bearings, see my article:
http://blog.discoverthat.co.uk/2017/08/filament-spool-spindle.html


==

Lower filament guide

I have found that the filament on my Ultimaker 2 flows best if it runs under the power adaptor socket.
This must make the angle better for the approach in to the feeder.

I have never been keen on that. Although I doubt there is much strain on the socket it does not look like a good idea.

To resolve that I have made a simple guide that fits under the lower lip of the Ultimaker case and serves the same function as the socket does to direct the filament.

It's taken a few experiments. I think I'm on mark 5 already.

To get that shape without a support structure it will be produced in two parts and joined with a dovetail.

I'm pleased with how well it positions the filament.

Unfortunately I'm fairly sure that the extra friction of using two guides causes under-extrusion when the reels of filament are part used. I'm back to cut lengths of filament for reliability.

==

Update: 2 August 2017

I have found that by running the filament past only the new guide works well. I have not tried any big models but my smaller tests have shown no sign of under-extrusion.

Test Tower

Based on that and the numerous guides and axles using bearings that others have designed, I am working on my own version with a pulley and skate bearings.

==

Download:

The 3D print test tower I've used above (stl)

==