Friday, 21 March 2008


Because my test objects are less warped while they are still attached to the polypropylene bed, I had the idea of filling them with something that sets hard to freeze them in that shape. That would also allow me to use a sparse fill pattern, which speeds up the FDM build time, but still get a strong object.

I needed something that was not too viscous so that it would flow in between the mesh of the fill pattern and would set hard.

Polyurethane was recommended to me because it has the consistency of milk before it sets and is strong enough to cast parts for Darwin. I bought some Smooth-Cast 300 which has a pot life of 3 minutes after it has been mixed, and cures in 15 minutes. I choose a fairly fast setting one because it gets hot while curing and I hoped it would soften the HDPE to relieve the stress. It only seems to get to about 50°C though so I don't think that it has much effect in that way.

This is the equipment I used :-

I know the internal volume of my objects pretty accurately so I measure out the required amount of plastic using separate labeled syringes for the two components. I mix it in a small pot before filling a third syringe to inject it. The syringes and pot are made out of polypropylene, which polyurethane does not stick to, so they can all be reused. I haven't found a way of unblocking the needles though.

I made a 50% filled object and drilled a hole the diameter of the needle in the middle that allowed the needle to go to the bottom. I also drilled a small riser hole at each end to let air out. Obviously, with cleverer software these holes could be made during the FDM phase.

The first attempt was a complete failure because the needle blocked when the object was only about 50% filled. Here is a cross section :-

For my second attempt I used a thicker needle, 1mm OD rather than 0.8mm :-

The object filled OK, but just as it became full the plastic in the needle set suddenly but I carried on pushing. The needle popped off the end of the syringe and PU sprayed all over the place. It was a good job I was wearing goggles and gloves but I should also have been wearing long sleeves, a mask and a hat! Fortunately PU does not stick to much, only untreated wood, skin and hair! Where it gets on your skin it burns slightly. Because it is transparent before it sets it is very hard to see where it has gone but when it sets it turns opaque white so it becomes obvious.

It actually sprayed around one quarter of the room. I even got some on my lips which I didn't notice until I tried eating.

What seems to happen is that if you subject the liquid plastic to pressure it accelerates the curing, which increases the temperature and pressure creating a positive feedback effect which makes it set suddenly in the needle. I only had two needles and they were now blocked so I did the remainder of my experiments using just the nozzle of the syringe into a bigger hole in the object.

The ideal solution is probably a very big needle that locks onto the syringe. It doesn't need to be sharp but the 45° slant at the end is handy because it stops the end being blocked if you press it against the bottom of the object.

I left the objects on the bed overnight to make sure the PU was fully cured even though it sets in 15 minutes. The first object I made had a 50% fill and warped 0.36mm compared with 0.47mm without the PU injection.

Thinking that 50% fill leaves the PU fairly weak, I did another test at 25% fill. That gave 0.24mm warp, the lowest figure I have achieved yet for this shape.

I also tried a 100 x 10 x 20 mm test with 20% fill ratio. That gave about 50% less warping compared to the version without PU.

A useful technique for reducing warping and reducing the build time of FDM objects. The main disadvantage is that FDM is one of the cleanest and safest fabrication techniques whereas injecting PU is messy and somewhat dangerous unless you wear protective gear.

I was disappointed not to get rid of the warping completely. Instead of alternating the horizontal and vertical fill patterns, several layers of one followed by several layers of the other might make the PU lattice stronger. Raising the PU to 50°C for a few hours is supposed to harden it further, so I could try removing the bed and putting it in a very low oven for a while. I have a Peltier effect 12V beer fridge which can be reversed and used as an oven, so that would be ideal.

Using a harder plastic like epoxy might work better but it may be too viscous to inject. I believe heating it reduces viscosity.

Reducing HDPE warping feels much like banging ones head against the wall so I will try PCL and ABS next for some light relief.


  1. I wonder what happen if you first printed a base surface that compensated for the expected warping and then printed your HDPE onto that.

    I'm setting up Tommelise 2.0 to be able to print in 3 dimensions to make that easier to do.

  2. Yes I think that would work in theory but you need to print a curved surface with more than filament height resolution. That would require controlling the extruder output along the trajectory as well as X,Y and Z.

    With my machine I could extrude a two layer raft and then mill it to a crown shape before extruding the object.

  3. "With my machine I could extrude a two layer raft and then mill it to a crown shape before extruding the object."

    That would be a damned good way to see if the idea would work at all.

  4. It looks like you're trying to overcome limitations of the material through engineering of the structure. Okay milling to a crown shape would probably cure the warp of a symmetrical shape, but surely, trying to predict how a complicated shape would warp is beyond the scope of the project. An FDM machine should just be able to lay down arbitrary solid structures.

  5. np,
    Yes FDM should be able to do arbitrary structures and it appears that indeed it can with ABS, PCL and PLA. There is a lot of incentive to get HDPE to work as well because it is cheap and easy to get hold of.

    It might be impossible but we have to experiment to be sure. Theoretically you can predict the warping with finite element analysis but the resulting correction needs support material and true 3D FDM. Theoretically my machine can do that, it's just whether it gets too hard in practice.

  6. It sounds like you are removing the shape from the raft, then overnight it warps more. Have you tried leaving it on the raft overnight and then removing it? Maybe it won't warp as much if it is still connected.

    On an earlier topic, you wondered why Cerastil was sold even though furnace cement is much cheaper and also stands high temperatures. I speculate that since Cerastil is designed for potting resistors, it may have higher thermal conductivity. If that is true to get the same thermal connection from the stainless tube to the heat sink, you would need more furnace cement than Cerastil. However, because furnace cement is so much cheaper you could put ten times as much to get the required thermal conductivity and the large amount of furnace cement would still be cheaper than the small amount of Cerastil.

  7. Yes I have left some samples overnight on the bed. I didn't notice them being less warped after I removed them, but I have not made any measurements to prove that.

    I agree that Cerastil is probably a better conductor of heat. With a less conductive cement you don't need more to get the same heat transfer, but less. I.e. you need to make the layer thinner which is difficult. However, the power used by the heater is not all that much, about 7-10W so it would not be a problem just putting more power in, although eventually the Cerastil would pay for itself in electricity.

  8. Hi!

    Before you completely give up with hdpe, I would like to propose the following idea:

    Try to heat the chopping board (bed), and fill the material with hot water.
    (for the heater you have posted this article:

    Thinking along this idea:
    If the material not fillable, build a wall around it, and fill that space with hot water.

    Please try this idea out, before you completely give up with HDPE.

    Thank you really much for your in-depth articles.

  9. My worry is that the warping is due to the upper layers contracting, while the raft is too dense to contract. By milling the raft, you're not stopping the shape from changing, you're just ensuring a flat bottom. The dimensions won't match up to the original layout.
    I think you need to stop the contraction, either by changing the properties of the extruded thread (thickness/tension/temperature) or engineering the path to include structures to allow the contraction without changing the outline.

  10. I wouldn't worry too much about anybody "giving up" on HDPE, khiraly. The rewards of getting HDPE working are simply too big for it to be given up with a LOT more work.

    It is, however, a difficult and frustrating plastic to work with and I can well understand Nophead's desire to try something else for a change. I certainly had that same feeling six months ago. Nophead's results, however, have encouraged me to give HDPE another hard look.

    That said, it has become obvious that you can't simply lay it down with your extruder like so much cake icing. Still, it's structural strength and higher temperature resistance make it too attractive to give up on so easily.

  11. khiraly,
    No I haven't given up and I will try the pcb_heater under the bed. Interesting idea to add a water bath rather than an oven. I may well try that as well thanks although that limits the temperature to 100C. Could use oil but lots of scope for mess!

    You can't stop the contraction, HDPE shrinks 2% between melting point and room temp. If it all cooled at the same time the dimensions could be adjusted by 2%. The problem is each molten layer welds to the solid layer beneath it and then tries to shrink 2% leaving the object under stress. The base holds it flat but when released the stress causes bending.

    I will study PCL next to see how it differs from HDPE in an attempt to understand why it does not warp. My first impressions are that it is fundamentally different at the solid, liquid phase change.

  12. khiraly,
    On seconds thoughts a water bath is not going to work on a moving table machine! It will have to wait till I make a Darwin.

  13. nophead,
    I think it depends of the table speed.
    So if it is slow enough, the water will rest inside the object;)

    I think its worth a try. (Im not proposing a long-time solution, but an experiment)

    Best regards,

  14. For the cement, I meant that more low conductivity cement would be needed to connect the stainless steel tube to the heat sink. For the nichrome wire heater, I assumed that you were using insulated wire and the cement was just there to hold the wire in place, I did not realize it was also for heat transfer.

    A way of increasing the thermal conductance of cement while not decreasing its electrical resistance would be to mix it with an electrically insulating material with high thermal conductance. A material I saw with those properties is aluminum oxide, which has a thermal conductivity of 30 W/m K, according to the table at:

    In grit form it is cheap, at 90$ for 25 kg, acording to the listing; 240 Grit White Aluminum Oxide - 50 lb Box -
    $89.00, at the bottom of:

    Unfortunately, even 240 grit is fairly coarse, 325 mesh is available at about $30 for 1 kg at:

    To reduce the creep of HDPE, you could try mixing it with marble dust, as described in Forrest's CAPA/marble dust post at:

  15. There are to related methods that may reduce or eliminate warpage - a higher temperature extrusion environment - I.E. put the extruder in an oven only 20C or so below the melting point. you could also try to do an anealing process - heat the finished produce almost to the melting point on a flat surface. This is interesting becuase you could attempt to recycle the warped test blocks, by fixing the warping at a later stage.

  16. What about printing 'islands', wait for those smaller patches to cool and shrink, and then connect the patches with relatively short bits of print? -Jasper

  17. Since I've seen how we stop overhanging PLA prints from curling up by directing a fan at the print, and how well so many things will stick to a film of PVA smeared onto glass, I've been wondering whether we can use both of those things in order to reduce HDPE's warping. I won't get around to testing it myself until I can get my hands on a benchtop 3mm filament extruder.