3D printed boats

Hello,

I'm new on this forum, so let me introduce myself.
For a quite long time, I have been experimenting with 3D printing in conjunction with composite manufacturing, at first only in pretty common way of printing small parts and/or patterns/moulds, but slowly working towards fully leveraging fast 3D printing as a viable production method for prototypes/small production runs of various boat hulls.

This is my latest creation, low volume, flat-decked paddling canoe, 405cm long and 46cm wide, with "recurve" keel-line for low stable position which is still powerful and ergonomic (feet below seat):



* 6kg of filament & 60 hrs of printing, it's essentially 3D printed core with optimal geometry and some interesting geometric structure (in the retrospect, lighter 4-5kg core would be totally sufficient)
* Skinned with GFRP from inside/outside to form a stiff sandwich, one layer of 320gsm biaxial fabric inside, one layer of 160gsm twill fabric outside (in the retrospect, 160gsm even inside would be fine)
* Wooden paulownia deck, skinned 160gsm from both sides
* White PU topcoat
* 10.3kg total weight (could easily be 7-8kg with lighter core and lighter carbon fabric instead of glass)
* Still very stiff and strong due to the very thick (up to 20mm) 3D printed core. I only did some basic back of the napkin type calculations for the whole boat, but when I place the bow/stern on supports and stand (80kg) in the middle, it's very solid with minimal deflection (it will be never subjected to such loads in normal paddling conditions)
* It still needs rudder with steering mechanism, but when I tested it on a lake, it behaved very good in water, with good speed/stability ratio, given the short waterline length.

I would love to get in touch with people interested in the technology/potential of rapid functional prototyping in similar way, feel free to ask any questions, comparisons with traditional DIY building methods like stitch&glue or strip-planking, etc.







Some photos form the print/laminate process:

 

Hi Jan, a very interesting project you have there! I am into 3D printing myself as well, so I have several questions for you.

1. What type of filament did you use?
2. What settings (perimeter lines, infil, etc.)?
3. Are you not worried about water intrusion into the core if the shell is damaged? When people build similar kinds of sandwitch hulls with foam, usual consensus is that no matter what you do, the foam will still end up soaked with water that you won't be able to get out. In this case, foam is substituted by air gaps inside the 3D print. Did you design in some channels in the infil to allow the water to flow to one end of the hull on order to let it out in case of hull breach?
4. What kind of Delta printer is that, and the extrusion system? 60 hours seems insanely fast for 6kg of filament.
5. What kind of joinery did you use to assemble 3D printed parts?
6. How much flex and sag was there in the 3D printed structure prior to applying glass? Did you find it difficult to keep it from breaking or splitting along layer lines when handling? I am also very surprised that a single layer of 160/320 gsm was enough. Seriously, what kind of filament were you using?
7. Would you mind sharing your CAD files, or at least some pictures with cross-sectional views? I am really interested in how you managed to make such a stiff structure with so little material.

 

Hi laukejas, very good questions, I will try to answer them:

1. It's plain PLA, annealed to raise the TG - I experimented with almost every FDM material available (except really exotic/hard-to-print materials like PEEK, PPS...) and PLA still wins for this purpose by a long shot, as it's by far stiffest FDM polymer with ~3.7GPA E modulus of elasticity compared to 2-2.2GPA of ABS/PETG, or even 2.5GPA of polycarbonate.
There are materials which promise much higher values, like special CF filled PA blends, but they rarely deliver on those values, and especially PA6/66 blends absorb moisture like crazy and load creep horribly afterwards.
For example I printed one small boat (2.7x0.44, for my 6y daughter) from much more expensive technical CPE filament and while it worked ok and once laminated it was very stiff as well, I had to use thicker extrusion lines due to the much lower E modulus (~1.8GPA) of the material.


2. It's impossible to print such things efficiently with standard slicer, I have my own special slicer generating optimised GCODE with minimal travel moves, next to no retractions and fully configurable infill types which is perpendicular to perimeter contours, unlike infill in normal slicer which is just fixed mesh cut by perimeters and the only way to "orient" it is by turning the model on the buildplate.
When I started I tried to use standard slicer with all kinds of tweaks, I tried surface-only mode in cura (popular in 3D printed RC airplane community), nothing worked properly and reliably, so I had to code by own slicer, now it's quite mature and I'm very happy with it. The structure used is basically warren-truss in cross-section, with sine modulated truss location between inner/outer skin in order to combat euler buckling. I did compressive and shear strength testing on stabilised samples of the core and they compared quite favourably with performance foams/honeycombs commonly used in composite construction, especially as the core grows thicker.
The boat was printed with 0.6mm extrusion width and 0.3mm layer height, 0.45-0.5mm would be probably fully sufficient, saving 1-1.5kg from overall core weight.

3. As you can see on the translucent core from CPE, there are sine modulated triangular channels which go along the whole piece, only separated by solid top/bottom where pieces are joined, so in case of a skin/core breach, it shouldn't be problematic to drain it out. Still. nomex cores work just fine in many higher end carbon boats and they are basically full of hollow hexagonal cells.

4. It's big www.spacedelta3d.com, my own printer design, scalable to big dimensions with very fast print-head movements, hot-end currently tops out at 30mm3/s, that's the main limit.

5. I used methylacrylate structural adhesive, although plain CA would be totally sufficient, once firmly laminated between FRP skins, that basically doesn't matter.

6. It was quite solid and I could easily lift and position it without any special care, no problems with twisting or splitting between layers.

7. My design program works directly parametric patches, or to be more specific, I specify inner/outer sandwich surface patch and the program computes the infill and directly generates gcode. I can of course share rendering of hull inner/outer contours (cubic NURBS in control lines & cross section) but unless you are specifically interested in that particular hull, there is nothing specific which makes it stiff, that's just the outer/inner shape, stiffness comes from the sine modulated warren-truss like infill structure.
Here is the rendering of sample cross-section from my slicer (it can do non-planar/variable width paths as well), hopefully the internal structure can be seen there:

 

I've roughly nothing to add intellectually... except to say that is awesome!

 

Thank you for answering my questions, this is seriously impressive, especially that you had to write your own slicer software for this, I know how incredibly difficult that can be. I have a few follow up questions based on what you wrote:

1. How did you anneal PLA parts so large, and without causing them to deform? I had annealed PLA in the past, but even if I kept temperature relatively low (60-70 C), there was still a significant dimensional deformation. Overall, PLA does sound like the best choice here, although I'm very surprised that you had no issues with layer separation, considering you only had one perimeter. Also, at 6kg, I dare not think how expensive it would be to use any of the CF-filled filaments...
2. Typically when slicer is unable to generate required infil, I design it myself in the CAD - what was the reason why this wouldn't work here? Is it because it would become impossible to control travel moves and retractions? Why was it so important to avoid them?

Your project made me seriously jealous! Since I'm designing my next boat right now (around 4.7 x 0.6), I would be considering the same approach, but since my printer is limited to 350 x 350 x 350 (Voron V2.4) I guess having to do the hull in such small pieces, the joinery and glueing would result in way too much excess weight. Still, damn nice project you have there!

 

I was wondering why the stern section is green? You are amazing I would not be able to laminate a skin on a carpet inside my house with a small drop cloth like that without getting resin all over the carpet Is there any off gassing running a 3d printer with your filaments?

 

Thanks, I have some other bolder ideas in my head including a small beach catamaran with wingsail and ultra-slender (length to beam ~ 20) paddling torpedo stabilised with long, slender bilgeboards (that may or may not work in practice)

@laukejas, to answer your questions:
1. The key is to anneal everything after it's laminated and use laminating epoxy with relatively high-tg + do it with slow ramp-up. I did some tests on smaller parts and it seems that even without separate annealing step epoxy exothermic reaction anneals the thin PLA skin at least to some degree, as those sample panels were able to withstand ~65c without any problems and deformations, which is well above tg of PLA (but below cured epoxy TG, even before proper post-cure cycle).

2. That may work when your infill is modelled by relatively thick lines created by multiple extrusion lines by slicer, because unnecessary travel-move and/or bit under extrusion is no big deal, but once you get to thin structures where everything is single wall, it's very unreliable -> the slicer basically sees everything as perimeter lines, often decides to do non-optimal jumps between them which in such big structures results in very long travel moves with bit of underextrusion, weak/unreliable connection between "infill" ribs and skins, etc. Not to mention print times, which are way, way lower when you basically have full control over paths. You are also able to force single print direction for any outward facing wall which result in absolutely gorgeous surface quality (makes laminating skins over core much easier compared to foam/wood).
Here is a good video of the process, you can see that the both infill and perimeter is realised as one long uninterrupted extrusion.

Voron is a very good printer capable of high print speeds and accelerations (especially with cf crossbeam), i think you should try it With ability to print very precise parts, joining and fairing the joints is no big deal.
Unfortunately our biggest printer tops at 50cm build diameter (but 130cm print height), otherwise I could print cores for your boat in just 4 pieces.

@Waterwitch The stern part is green just because i ran out of the grey filament and had to change mid-print, it's the same plastic type. I run all my 3D printers in separate room in my office spaces, that's also where i laminated it (I was alone there) with full PPE suite and good venting afterwards, i wouldn't do that at home

 

Amazing stuff, man, thank you for answering my questions. The video you shared clearly shows the benefits of your slicer. By the way, is it open source? It would be really cool to try it.
I do have some ideas of how to (big maybe) do it with standard slicer by importing perimeter and infil as separate objects into the slicer, and applying different settings to it. I'll try it when I get home.
About the annealing, I didn't quite get it, you annealed the entire hull after applying fiberglass and epoxy? Where did you find an oven this big?

One last question... I understood how you glued the separate pieces together, but I am not quite sure how you aligned them together, and how large the gluing surfaces were. Did you design in some aligning features, and maybe something to increase that bonding area? Perhaps you could share a few pics of two ends that fit together to be glued?

 

Interesting work, congratulations.
I suppose that all the possible problems inherent to the mechanical properties of a material that, in some aspects, is different from the traditional one have been solved and if these properties have been kept the same throughout the length.
On the other hand, when observing the process, I wonder if it would not be a better idea to use the 3D printer to build only the mold and on it, apply the normal technique of lamination with a male mold.
Thank you for your time.

 

@laukejas

1. It used to be open-source, but it's not anymore, recently I did a big refactor making it much more generic and usable and closed-sourced it. I would prefer to keep it open source, but I'm trying to turn the whole idea into start-up and raise-money for it, and every potential investor I spoke with freaked-out when I mentioned it was open source. I think it's stupid and I totally love open-source software, I contributed to multiple projects, including 2 PRs related to delta printers which were merged into Klipper firmware, but I need to raise money to push this whole idea further.
Modelling every extrusion line as separate STL could work, it's still very tricky to model enclosed volume with right thickness so the slicer always choses to represent it with one continuous extrusion line, doesn't interrupt/omit the lines, etc. Cura has experimental "surface only" mode which helps with this a lot (you basically don't need to model enclosed volume), but it's very buggy and when I raised one issue related to mode 2-3 years ago on github, it was basically dismissed as "not a priority" (I guess not many people us it) and closed without resolution.
At the same time, modelling couple of straight ribs manually in CAD tool is no big deal, but modelling warren truss following the contours of skin while being sine modulated in height, I really wouldn't want to do it manually...
2. Yes, it was annealed once laminated, I have a long, narrow (4.2x0.5x0.5m) low temperature "oven" made from polystyrene plates with plywood bottom and silicone heating wire as a heat source, hooked to simple PID controller, it's very basic and only gets to 60-70C (depending on outer temperature), but it's enough for annealing PLA over long time period.
3. Regarding alignment, the panels have full bottom/top and are joined with simple butt joint, I just ensure that inner/outer surface align perfectly, but it would be better to model alignment holes there and use short wooden/cfrp dowels there, something to work on.

@TANSL
Yes, I tested core samples quite extensively and they compare very good to traditional core materials, even not taking into account advantages like being able to use core with much greater thickness, smoothly varying core thickness/density, etc, with more development, it will be even better.
Regarding mold making, yes, it's an option, but I was never fan of male molds for hull making - the greatest advantage of molding is easily reproducible super smooth surface finish, which means female molds for ship/aircraft/vehicle hulls.
Directly printing female mold with supporting structure & flange doesn't make much sense IMHO, as sanding/polishing concave surfaces is much more difficult harder then convex ones, not to mention that 3D printed surface would never be such hardy and reliable as dedicated tooling gelcoat.
That leaves us with printing dimensional pattern and taking mold from it with traditional methods - the process I presented is just making the dimensional pattern not only dimensionally accurate, but functional as well, enabling rapid, relatively low-cost testing/iteration without need to make expensive female mold everytime something changes.
The greatest added value lies in the fact, that you can actually throughly test the hull in real conditions, opposed to only CFD simulations (which are great, but no substitute for real-life testing), iterate fast and once you are satisfied with the shape, take the surface to class-A finish and take female mold from it using traditional methods for economic higher-volume production.

 

Thank you for your answers, @Jan Herich, I'm not very convinced about them but forget it. It is not about convincing me but about making a hull that is resistant and safe, and cheap. I have never spoken, by the way, of a female mold.
Good luck, then, with your project. I hope it will be a resounding success and that we will soon see this procedure generalized, at least for small canoes.

 

I have seen some attempts at using 3D prints as molds, but it hardly seems like a scalable solution because of how much prep work is required to achieve surface finish acceptable for a mold. However, when using it as a core material like Jan did here, this issue is not present (one might even argue that the rough surface of 3D print can improve mechanical bonding), and it may even have many advantages over traditional core materials for a DIY boat builder, such as:
1. Traditional core materials need molds to be put into shape, just like the inner/outer "skin" material. 3D printed core already comes in shape, so it's a mold by itself. No other molds, no jigs, no strongbacks, nothing really necessary, just prep the surface post printing, coat it with GFRP / CFRP and you have a hull.
2. Any kind of shape can be printed, opening up possibilities for some really interesting geometries that might not be doable with traditional female molds because of negative pull angles and so on.
3. Most core materials (to my knowledge, I might be wrong here) absorb water to some degree, so if there is is penetrative damage, it might be impossible to dry it out. Meanwhile, this mostly hollow 3D printed core would allow intruded water to flow along the infil like Jan said, and a collection area with a plug could be designed at a transom to evacuate the water. Or, since such occurances are so rare, just drill a small hole for the water to drain and later plug it with epoxy.
4. The flexibility of shapes that can be printed would allow for making other boat parts as well. How about centerboards, rudders, perhaps even spars?

I still have my doubts on the strength of the plastic. I've printed with PLA a lot, and even if printed at higher temperatures and annealed, layer adhesion is still a major strength issue. It clearly wasn't an issue with a canoe hull, but what about sailing dinghy hull, which is subjected to twist and bending? Since layer lines happen to go longitudinally along the hull (like stations), bending load would place most stress on these layer lines. That goes double for trying to build centerboards, rudders and especially spars with this method.
However, if we assume that this printed core is only really responsible for handling compressive loads, and everything else is relied on the GFRP / CFRP skin, then perhaps layer adhesion strength issue can be (almost) completely mitigated, since these materials are far stiffer than PLA, therefore any excessive bending force would first break the skin rather than separate print layer lines, meaning it layer adhesion doesn't really matter. I think I might make some test prints, coat them with CFRP and load them to failure to test this theory. If this holds true, then only the compressive strength of the core really matters, and everything else is taken up by the skin. Since 3D printed core can offset the skin from the neutral bend axis by such a huge amount, adding several more layers of CFRP would increase strength in the most optimal location in terms of weight.
Jan, I would really love to hear your thoughts on this. Perhaps you already made some similar experiments?

You also mentioned that you made some simulations to calculate the expected strength of this PLA core + CFRP combo. I would really love to hear how you made these simulations. Since 3D printed parts are highly orthotopic, and even disregarding layer adhesion, the specific print line layout can highly influence stiffness and strength, I always assumed it would be near-impossible to simulate with any degree of certainty. And to add on top of that, making a composite material of such a structure with CFRP coating, which is also quite a difficult material to simulate, seems to make it even more complex and unpredictable. Jan, can you share any of this info? I seriously think you are onto something with this project, but we definitely need to work out what are the limitations of this new technology.

P.S. It would also be interesting to know if epoxy binds chemically to PLA (or any other common 3D printed plastics like PETG, PC, ABS), or just mechanically. As far as I've tried coating PLA parts with epoxy, the bond seems to be really strong, but then again, I would always scuff up the surface with 80-120 grit, so I'm not sure really.

 

Quick addition: I found a very interesting paper on the subject: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6401912/pdf/polymers-10-01262.pdf It's a study of 3D printed materials coated with composites and an investigation of the resultant material properties. There are a few interesting moments I'd like to point out:

1. In section 2.4.1, there is a consistent pattern that 2 layers of composite coating on top of 3D printed core is significantly weaker than 1 layer, but 3 layers is stronger than either of the two. There seems to be something weird going on with 2 layers. I am still searching through this paper for an explanation.
2. In page 7, they state:

It was noticed that the failure of the sandwich CFRP/ABS/CFRP started at the CFR layers
followed by the failure in the core material. This because the ductility of the CFRP is less than the ABS
material.

I interpret this as a confirmation of what I wrote in my previous post, that layer adhesion issue inherent to 3D printing might matter not matter at all with composite coating, so perhaps the only question of the applicability of 3D printed materials is the compressive strength per weight compared to other core materials. Feel free to correct me.

 

I don't think it is like this - if I understood you right. It is just a composite material (sandwich material) 3D printed or not. The typical stress for a part of a hull is bending stress and it is causing shear stress. To take the shear forces a strong adhesion of outer layers and core material is neccessary.

 

Hi @laukejas
Regarding the paper you posted, I found the whole concept and methodology super strange:
1. They printed dogbone samples which are used for tensile strength testing
2. They laminated them to form sandwich structures and and measured ultimate tensile strength/E-modulus -> as the cfrp skins have much absolute higher tensile strength & E modulus (even taking into account significantly lower cross-section of cfrp compared to core material), the core is basically totally irrelevant (as long as its strength/E-modulus are lower in absolute terms compared to skins), and you will get the same/similar values from the test machine even with no core whatsoever, just 2 skins bonded together.
The only influence of the core is on specific strength/E-modulus of the part because the core changes the part weight/density.

That's not how sandwich structures are usually tested, testing bending strength/E-modulus would be much more insightful as that's the whole point of sandwich structure.
Not to mention there is no data on infill type used, print orientation, core thickness...

When I tested the cores, I first printed small stabilised samples and tested their specific strength in compression/shear - I compared those values with datasheets for foam/honeycomb/end-grain balsa etc.
I didn't have access to proper expensive testing machine, so I unfortunately couldn't really test E-modulus, I just designed simple fixtures and progressively loaded weights till it failed.
Next I created rectangular laminated samples and tested those with 3-point bending test, already measuring displacement in the middle via dial indicator.
Every failure there was in buckling mode, either in the skin loaded in compression or core rib loaded in compression, never delamination of core between layers or skin/core - that was the primary motivator for sine modulated structure, in order to raise euler critical load.

While FDM prints are always weaker in the Z direction, it's not really that bad and depends a lot on material/cooling -> PLA/PETG with minimal/no-cooling exhibits Z strength of 70-80% of the optimal direction, which is kind of OK.

Regarding chemical bonding of epoxy to plastic surface, I doubt that, I always scratch the surface with ~80grit sandpaper for good mechanical bond, low-surface energy materials like nylons would probably benefit from things like flame treatment as well.