I’ve a few projects where I need a flexible rubber-like 3D print including some static pump seals for Yurtle’s fish tank pump, a new watch band for a Seiko watch, some gaskets for my Amigo test cases, and some flexible moulds for turtle calcium blocks.

Pump Seals

The most pressing of these projects are the new static seals Yurtle’s canister tank pump and filter.  Sure I can buy some from over-seas but the replacement parts are expensive, even before I tack on freight, packaging and import tax.

Although the wall thickness of the OEM part is quite thin it is geometrically simple, small and self supporting.  I recommend using a relatively small and simple model when starting out with a new filament.

Figure 1.  Original Eheim Pump Tray Seal

There are a number of rubber 3D printer 1.75 mm diameter filaments on the market but I thought I’d try NinjaFlex.

Measuring the model from the original seals was a bit tricky.  Two of the original seals have hardened, are distorted and torn.  However one was still serviceable.  I also have all of the seating and mating components for developing the model.  Measuring rubber parts with precision is difficult as it compresses and flexes under compression from a micrometer or callipers.  This was always going to be an iterative design process.

Figure 2.  Model

The original part has a typical wall thickness of about 1 mm but I want the walls to be at least three layers thick (1.05 mm minimum for an 0.35 extruder nozzle).  The seal doesn’t need to be perfect which is useful because 3D prints are porous, and thin 3D print walls are more so.

The original seals might also be improved by increasing their seating length.

With a prototype completed in Autodesk Inventor and sliced using Simply3D the model was inspected to ensure that all layers were at least three layers thick.

Figure 3.  Walls at Least Three Layers Thick

Now for some NinjaFlex print parameters (speeds, feeds, temperatures, retraction, and so forth) for my M2.  Rather than start from scratch I used the general guidance on the NinjaFlex website and took the most appropriate parameters from other M2 users that had posted their setting on the web.  Note that the M2 has a direct drive extruder which is recommended for NinjaFlex.

Next I loaded the NinjaFlex filament.  First up I heated the extruder to 220°C and retracted the existing PLA filament, followed by extrusion until clear under Protoface manual control.  Then I fed the NinjaFlex into the extruder which was a bit tricky because the stuff is so flexible.  However it fed directly into the extruder and extruded without adjustment of the filament compression screw.  Note that some folk report that repeated high extrusion can cause problems due to inadequate filament heating so wait a few seconds between the default Pronoface 10 mm extrusions.

Time to print.  The first few print trials were a disaster.  My second extruder head was ripping the first layer skirt perimeter off the bed.  The problem was not the extruder alignment, but a small blob of old PLA filament on the idle extruder.  I heated the extruder up to 215°C and set it to cool while wiping the tip with a clean paper towel.

Finally I was printing!  The finished print was a bit rough with some wall porosity and the model needed some adjustment to better fit the pump parts, but the print parameters are close.

Figure 4.  Initial Print
Close - but no cigar.

When adjusting print parameters try and be systematic.  If you change a whole bunch of stuff at one time then you’ll have no idea what was useful and what wasn’t.  This includes the model!  Changes shouldn’t be too extreme.  Ten degrees is a relatively large temperature change within a manufacturer’s recommended band of about 20°C.

Re-read the resin manufacturer’s specifications as a guide, look at the printed part and think about what might be causing the defects, take some measurements (printed filament width and layer height),  think about bed adhesion and weaknesses with the model (unsupported print angles and the like).  Every printer and print environment is different.  Your printer may have problems even though ‘the other dude’ got great prints with a particular set of parameters on exactly the same model of printer.

Figure 5.  Rough Prints Through Parameter Adjustments
 (Too hot left and too cold rght.)

Sure, you’re going to waste some filament in this parameter tweaking process but you should eventually achieve a good useable print and end up with parameters that should be a basis for other models (but note that every model is different and might be refined by improved print parameter settings).

Figure 6.  Final Seals - Nice!

Some general notes on printing with NinjaFlex.

• Print slow (say 500 mm/min) with X and Y non-printing movement as fast as possible.
   
• Ensure that thin part walls are at least three layers thick (two solid outlines and overlapping infill).  If you don’t achieve this then you can expect that your printed part will tear between layers under flex and/or tension.   
   
• Try and ensure at least three solid top and bottom layers.
   
• If the extruder is too hot or too cold the print will produce rough walls with visible  penetrations.  For my M2 printer 215°C is too cold and 240°C is too hot.  My optimum extruder temperature was 225°C with the print cooling fan off for the first layer and on at 50% for the remainder of the print.
   
• A bed temperature of 50°C provided good bed adhesion.
   
• Retraction is of limited value as NinjaFlex resin will simply flex and string.
   
• Initial model skirts are useful to ensure sound extrusion.

The printed part will be flexible which can cause interlayer bond issues for thin unsupported parts as the printed portion of the part may flex during printing.  NinjaFlex makes excellent support material for NinjaFlex although it can take some effort to separate the support material because it adheres to the base part very securely.  As an alternative you might try ABS or a proprietary soluble support material.

After refining the print parameters and adjusting the model I was making good seals that fit well and are functional.  The surface finish of the printed parts is great, although there were some minor internal strings from movement between print regions within a layer.  The resulting part is very flexible and strong.

Where you need a great surface finish and good inter-layer adhesion vapour sealing can work really well.  I am currently trying concentrated sulphuric acid (a relatively common reagent) but NinjaFlex is soluble in many concentrated mineral acids and some rather nasty solvents.  While NinjaFlex is also soluble in concentrated nitric acid this also causes the filament to swell distorting your print.  So avoid nitric acid.  If you want some other solvents then contact NinjaFlex or go on line and look up non-swelling solvents for polyurethane.

You can download my M2 print parameters for NinjaFlex sliced with Simply3D on the following link, but note that these are for the second extruder on a dual extruder M2 printer.  Your hardware and firmware are likely to be different.  Simply import the FFF profile into Simply3D and make whatever adjustments you consider are appropriate for your printer.

NinjaFlex.FFF

For a more complex NinjaFlex Model link to my replacement Seiko watch band.

Solvent Smoothing Experiments

I’ve started on my vapour smoothing experiments with NinjaFlex.  This started with looking up solvents for polyurethane to identify those that were likely to be effective and would not cause swelling. The was reduced to a rather long ‘short list’ of common reagents that I have on hand.

We don’t actually want our plastic parts to rapidly degrade in common solvents because this would render them pretty much useless for many applications.

Rather than mess about with vapour for now (which is a very effective way of getting a part coated in a thin solvent film) I simply put a piece of the failed watch band print into the solvent at 20°C for up to 60 seconds, washed (and in some cases neutralized) the test piece and examined it for surface finish, delamination, loss of flex and tensile strength.

Concentrated sulphuric acid was by far the best solvent tested to date but it instantly turns the NijnaFlex test piece from black to grey - and is slowly darkens again after several days.  This won’t do at all.  Literature suggests that Nitric Acid will cause significant swelling so this won’t be on the test schedule for now.

Figure 7.  Concentrated Sulphuric Acid Turns NinjaFlex Grey
(Exposure from left to right: 0, 30 and 60 seconds)

Concentrated Phosphoric Acid had little if any action - perhaps a slight dulling of the surface.  Many organic solvents were also completely ineffective including alcohols (Methanol, Ethanol, Propanol), Mineral Turpentine and gasoline (petrol).

Acetone, Ethyl Acetate and Methyl Ethyl Ketone caused the NinjaFlex to rapidly swell and delaminate between layers.  Oops!, these won’t do either.  I expect similar results from all polar halogenated hydrocarbons.

So I have some more solvent experimentation to complete.  Further literature research suggests that TetraHydroFuran (THF) is the solvent of choice for this application.  I’ll be ordering some in the next few weeks.

There is also another possibility that may be effective for surface sealing - heat treatment using hot air or a precision oven.  Obviously we don’t want to be getting the band too hot or it will either turn into a molten blob, char or ignite.  Hot air at 230°C was effective at melting the test piece but this was very difficult to control.  Thin sections and edges heated very rapidly and started to flow before the bulk surface.  When the bulk surface had begun to melt the entire part was hot and started to deform.  Maybe I can used hot air for a cosmetic touch-up, but my experience is that its not much chop for surface smoothing and sealing of the entire part.

So You Just Want a Replacement Seiko Band?

After a week of 24/7 wear the new watch band is as good as first fitted.  It remains as comfortable as the OEM band with no stretch and no skin irritation.  The only down side is with water immersion as the open print absorbs a small amount of water.  This will be corrected once I sort out the solvent sealing.  If you want a replacement band for your watch then please email me.  I figure that I can make these for you (two for US$20) plus postage and packaging.  The cost will increase slightly once I sort out the solvent sealing.

I can probably do other replacement bands that are not commercially available as well.  Please feel free to ask if I can help.

Plaster of Paris Moulds

I’ve been making turtle calcium blocks every month for over a year now and Yurtle loves them.  But the current mould for the Plaster of Paris slurry makes releasing the block in one piece a bit tricky, particularly if the block has set solid.  NinjaFlex should be ideal for Plaster of Paris moulds.

I remade the mould model from the original as a 3 mm thick shell with support structures that form a stand to keep the mould level.  These also provide grip points for stretching the mould to release the block.  The part was printed with NinjaFlex support structures to 45 degrees.  The part has a relatively long print time of about 7.5 hours so, after checking that the first few layers were sound, I left it printing over night.

Figure 8.  NinjaFlex Plaster of Paris Mould

The new mould is sooo much better than the original PLA mould.  Even though I haven’t sorted out NinjaFlex solvent smoothing, the block releases easily without damage (even without using an oil release agent), and the definition is just fine.  Whoot!

Figure 9.  Truly Flexible Mould Makes for Easy Release

Vapour Smoothing

I have continued on with my solvent smoothing experiments.  As noted above, TetraHydroFuran (THF) is identified in literature as being useful for 3D print smoothing of polyurethane based filament.

My first experiment involved solvent dipping but this caused rapid delamination.  That’s not good, but after some thought the reason for this has become apparent.  3D prints are porous (get used to it) and solvent dipping allows a mobile solvent to rapidly migrate between layers through the part resulting in delamination.

However with vapour smoothing a thin layer of solvent condenses on the cold outer surface of the part and does not migrate appreciably between layers.  Vapour smoothing with THF works well.  THF has a boiling point of 68°C (suitable for water bath heating) and is arguably no more of a fire or health hazard than petrol.  While inhaling solvent fumes is not recommended for any solvent, THF has a fruity ester smell which is somewhat more pleasant than alternatives.  Glasses are still required for safe handling, but without the inherent dangers of concentrated mineral acids.

For the vapour smoothing experiments I poured about 10 ml of THF into a 200 ml beaker and placed this in a tray of boiling water.  A second 400 ml beaker containing cold water was placed on top of the 200 ml beaker to reduce THF vapour loss.

The test piece was half of my second failed watch strap.  I cut it down its length which gives me a reference (status quo) against which to compare the treated half.

Once the THF condensation layer was apparent at about half the height of the 200 ml beaker I immersed the test piece in the vapour using a metal support wire.  Immersion was just long enough to observe THF condensing on the part as a shiny film (but without drops forming).  The part was withdrawn and allowed to dry in open air.

Don’t touch the part until it is thoroughly dry as the surface will be soft and easily takes finger prints.  Once dry the part can be re-treated if the surface isn’t quite as shiny and smooth as you’d like.  The surface dries hard with improved sealing between layers and filaments apparent.  There was no apparent loss of strength or tendency to delaminate.

Figure 10.  Vapour Smoothing (Upper Piece Treated)

Figure 11.  Vapour Smoothing (Upper Piece Treated)

Note that the remaining solvent can be stored for future use, but don’t pour in back into the original container as this risks contamination, make sure the bottle is appropriate for the solvent, and make sure it is clearly labelled, dated and stored.

These results suggests that Acetone, Ethyl Acetate and Methyl Ethyl Ketone and polar halogenated hydrocarbons should also be effective for vapour smoothing (as opposed to solvent immersion which my experiments show causes delamination).  These solvents are also likely to be more readily available than THF.