Didn't want to show that hours of adjusting settings could be beaten by yeeting more material out of the nozzle 😆. The weight comparison is nice. But unless it's flying I never cared.
What an awesome technique! It seems useful for being super efficient with your materials if using an expensive filament or printing a lot of copies of a model, or just for adding a visual detail like the dual color versions you printed. Also, thanks for confirming my intuition that adding perimeters is generally more effective than increasing infill, that little test alone made this worth watching :)
On the chance you read through new comments on old videos. A video highlighting whether or not going through that process with the previously strengthened completed hook to see if it will improve it any further.
What we need is a finite mesh simulation that outputs a weighted mesh that gives the stresses as a proportion of maximum. Then we need slicer integration that will allow you to create gradient infill according to a curve. That way we get better fading between infill strengths so that the load is distributed more evenly and we can fine tune the process at the end instead of restarting each time. Distributing the load would probably help a ton because it looked like some of the parts began to fail when the internally dense region began separating from the rest of the part. If it could be faded more evenly to distribute load, possibly with custom infill shapes to spread forces into the whole of the part, I bet the part could be made even stronger at the same weight. Effectively, I think what a large goal should be for the 3D printing community is for us to move away from static infill patterns to infill that is created procedurally for each part to maximize load distribution. I'm imagining a version of infill that has almost that tree-like branching that moves from a thick area near the main stress in a direction that will support the part and then spreading and splitting outward to spread out that load. 3D printing could allow us to, in a way, create parts that have fully connected fiber reinforcement aligned to maximize strength.
I'd imagine optimizing the filament paths is probably harder than just density. Although optimized paths will likely perform notably better than pure density. Especially since strength in various orientations differs non linearly for orientation and density. However, there's already ongoing research into fiber orientation optimization for high end carbon fibre structures... just need some of that for 3d print software.
Spot on. The vertical gradient between 100% infill and 10% infill was a stress riser. Interpolate over some range at the boundary and it'll be stronger.
Building off of this concept, I imagine it would be helpful to have different types of material within the model. PLA and flexible in the same model might provide many advantages.
I think I'd like to see more on the shape optimization using multiple materials. In the video you show the one printer printing the hook with the optimized infill in a different color. Instead of another color use a stronger material. So for the base hook body maintain the standard PLA and then for the reinforcements use one of the PLA filaments that has adders in it, or print that section with say ABS or Nylon. I am not sure how the two material bonding would react but it would allow you to use a cheaper material for the base part and then a stronger but pricier material for the areas that need the reinforcements. Similar to some injected molded parts where there are metal reinforcements in critical sections of the model (one thing that comes to mind is metal reinforced feed lips on polymer AKM magazines). Great work as always sir, from one engineer to the other, I love the way you structure these videos and tests.
Good idea, but rather than different plastics which print at drastically different temps I would suggest reinforced vs non-reinforced plastics. I.e. use standard cheap pla for most of the part, but on the high-stress regions use carbon fiber/ Kevlar/ cellulose reinforced pla.
@eugenioLU91 I haven't seen that, however I question whether you're thinking of structural or decorative pla. Either way it's a better place to start than attempting to print abs into pla (which will melt straight through the surrounding pla producing little more than plastic blobs).
@@robertasumendi I haven't had opportunity to use much of either, however the ability to mix materials seems like a promising niche for additive manufacturing. Personally I've been spending more time researching the potential of printing protoboards easier (as an ee that is more aligned with my use case)
It's obvious the reason why the "full optimized" design was strongest is because it delaminated before it could fail in tension. That allowed the structure to deform and relieve some stress.
Very interesting. I cant help but wonder how this compares to 100% fill of the whole hook. Also, then reduce the thickness or width of the hook to get back to the same weight at 100% fill.
You are my gurú! i work in a 3D printing laboratory here in México ,so whenever you come to Yucatan you have a room reserved in my home! you inspire us buddy!
Hey Stefan. From my understanding, the main factor that you want to keep in mind is the contact surface area in the plane perpendicular to the force vector. Hence you want to optimise force/(surface area). Solid infill certainly increases the contact surface area. The other factor that you might want to consider is the print direction for the infill in those place, as you have previously shown that it matters a lot (i.e. layer adhesion < material strength).
What an interesting group of thought/structural experiments. I'll have to keep those tricks in mind. One thing I've done for parts that I need to tap (When we make samples of castings at work) I cut a bunch of .02mm lines radially around the hole so that the printer does a zigzag pattern around the hole to be tapped. This way I can force plenty of solid plastic around the hole to finish drill and tap without having to up the perimeters everywhere else.
The displacement sharp edge you found is because something called triaxial state of stress. The hooks are failing like simple engeniering materials because the rupture is always gonna start in the inside (like a relatively ductile material). It happens because there are not so much stress concentrators (like the sharp edges of the infill) and the deformation is not possible because the surrounding infill is dense, blocking it. I'm sorry because the broken English. You can find more research material on this searching the subject fracture mechanics.
As someone familiar with the use of FEA this is an interesting topic and workflow. It would be interesting to see how the parts reacted when subjected to loads that it wasn't optimized for compared to the standard infill parts. At least with a hook it's fairly easy to know how the part will be loaded, but sometimes parts can be subjected to loads that you may not suspect, hence the need for safety factors. However, I'm sure this consideration is probably overkill when you're worried about hobby projects, but something to keep in mind nonetheless.
Yeah, but that would be really hard. The case where quadrocopters break usually is in a crash, so in my opinion you would have to optimize for this case. Using this method with FEM simulation, you only optimize for ONE case, usually where the force is one way and constant... Also in flight, the forces vary, if you change direction, because of inertia (Trägheit). So you would have to combine optimized infills of many cases. That might be interesting.
Design 2 parts in the sketch model and set infill for each separate part then combine them before slicing is a quick workaround and less time. once you see the warning infusion design a simple part to cover that area and import both models then set different infill for each presto a simple way to do it without the complication I've done this for years on mech parts.
Stefan - the one nice thing about the partial fail before full fail is if you had this in an industrial situation - you could in theory monitor and take action for that 'warning sign' before deciding the part was too close to it's tolerance limit before actual failure. On the other hand, it means you have to leave a larger 'reserve' in strength vs. a 100% infilled part which may make the savings (weight, cost, performance) trade off not worth it.
To create acutual fibers, you could generate a 3d grid of fibers and then use the combine/intersect tool to just get the parts of the grid, that intersect the simulation mesh. Then you could use this in other slicers, as it is baked into the model.
Ive done stuff like this in the past with Fusion. Never tested like you did but you did a great job explaining the process along the way. As for the gaps on concentric infill, what about using a larger nozzle diameter and low layer heights? This will effectively reduce any voids in the part. Also in fusion you can modify the shape optimization criteria for a higher target mass so you can get an optimization that is all connected. Which will likely reduce voids going from solid reinforcement to infilled spaces.
A very simple way of adding the smart infill :). Great work!!! An idea I came up with, but lack of time has stopped me :(. I was looking into using a basic fractal branching set of connected rods like bone. So the sides can spread the stress more evenly. I also think it’d help if the part was PLA and baked. As to what part you should try, I vote for the quad copter ;)
You're a smart young man Stefan. I'm new to 3d, just having gotten my first printer last week, but have been watching your shows for quite some time. I Love your content man!!!
Wow, imagine implementing this on 3d printed metals to make parts lighter but still as strong as a solid part(or stronger since there's room for strain relief within the structure)
There is a number of industrial software available in this regard. The keyword here is topology optimization. But sadly this really is not for the DIY community with prices starting from 10 k to well over 50 k € per year. withinlab.com/software/new_index.php www.ntopology.com www.materialise.com/en/software/3-matic ... the list goes on
16:40 Das was du meinst ist ein schön zu sehender ausgeprägter Bereich nach der Streckgrenze (siehe Lüdersdehnung). Das tritt vor allem bei zähen Kunststoffen (und niedrig legierten Stählen) auf. Dieses Verhalten ist bei zähen Kunststoffen stark von der Temperatur und Belastungszeit aber auch durch Kerbwirkungen in der Typologie abhängig. Bis zu diesem Punkt der Belastung ist die Dehnung komplett elastisch, es tritt also nach wegfallen der Last bis zu diesem Punkt keine plastische Verformung auf.
At 13:40 you can see that the hole in the bottom left that was supposed to be normal support infill is entirely 100%, though an isolated hole in the center in the modeling was not intended to be so.
This is outstanding! GREAT work. Thanks for sharing. I've worked with topology optimization for my master degree, studying the application of top-opt techniques in the early phases of an industrial design project (how to deal with style, ergonomy and so on…). This is such a clever application!
This is really interesting. We've been using Slic3r's volume-based variable infill to solve issues with part strength so I'll give this a go and test the relative differences in strength. Thanks!
since forces here are somewhat in one direction, infill could be done in that direction as well, in all cases (vertical). ie. horizontal density 20lpi, vertical 7 lpi. so that more vertical strands are present. also, continuous strands have more strength so they could be as solid as possible from top to botton, in an S, J or whatever shape needed.
Great Idea! I used Simplify3D and 2 combined "layer by layer" print processes (The main model with a sparse infill and the optimized meshbody .obj at 100%) with a similar result. In S3D one can just import the "optimized" .obj file as a "modifier" file.
That's the most interesting addition I've seen since the last few years in 3d printing. It would be really nice if this feature would be added directly to the slicing softwares. It would be nice to see the increase of streingth of a 3d printed object combining the effect of optimized infill and copper electroplating.
What you observed in your final graph is not a release of internal stress as you hypothesis. In engineering we call it the elastic and plastic stress strain curve. The initial part is elastic until you reach the drop meaning it will be permanently deformed. The plastic is just the state of permanent deforestation until breaking.
I know this is a stretch but I wonder if it is possible to use the pathing and infill data from the slicer as real geometries inside of the simulation tool (as internal cylinders perhaps?) to produce much more realistic simulation models.
That was a significant strength increase, fascinating topic. Also a great way to optimize. Most strength for minimum material. What I'm curious to know is what would the strength test be if you made the entire part with 100 percent infill. Feel like this would give you an upper limit on what's possible in terms of strength with the given material, and provide you with an idea of how close you can get to a theoretical maximum strength.
The hooks that had two bumps on the end of the load curve exhibited what is known as lack of composite action; i.e., the inner and outer structures did not act as a unified assembly. It looked like the outer skin failed under flexure followed by the in-fill core failing under buckling or shear.
It would be interesting to see how the smart infill pieces strength compares with itself, after annealing them. Great video and awesome technique, hope to see (and test) more of this!
It is great to see that topology optimization is finally becoming popular (it exists since 90ies). Great application! Some comments: Ansys has also topology optimization integrated (it had it already 10 years ago, but it seems they fogot it) There is also topology algorithms to optimize the orientation of fibers like CAIO, this might be usefull to align the printing orinetation. Unfortunately the software developers are a little bit slow to implement it. There is also latice structure optimization. As far as I know only altair offers software for this and it probaly can't concidder the anisotropy of fdm.
I think making sure the modifier has a good connection to top/bottom/perimeters would be in of high importance. Having the modifier connected by standard infill to the shell will probably have lower performance. Perhaps using the shape optimization mesh as a guide to hand model the modifier would yield the best results.
This is an amazing video!!! I really appreciate the work you into it and it's a testament to how proper engineering techniques can yield great results. At a bare minimum, any part I print that requires strength will likely be the 5 perimeters/10% infill, which seems to be a nice labor/strength improvement tradeoff.
Hallo Stefan! Als Bauingenier kann ich dir sagen dass du wirklich tolle Arbeit machst. Eines möchte ich dir aber bezüglich der "mesh-size" mitgeben: Spannungsspitzen können zum Teil sehr hoch sein und auch dort auftauchen, wo dein Netz gerade keinen Knoten aufweist. Im Bauwesen ist das meist bei Ausnehmungen , Einsprünge und dort, wo Stützen sind. Hier würde ich das Netz etwas kleiner ansetzen um wirklich die Spannungsspitzen zu bekommen und zu handeln. Was auch eine Rolle Spielen kann: das netz geht von einem Festpunkt aus und kann daher andere Ergebnisse bieten, je nachdem wie das Bauteil verschoben wird, da andere Knoten berechnet werden. In wiefern das beim 3d-Druck mit anderen Problemen (layeradhesion usw) eine Rolle Spielt lass ich mal dahin gestellt ;). Lg Thomas
This video was absolutely fascinating... You could use this method for a lot of applications. I got a little confused towards the end of the results, maybe you rushed through them too fast, or I'm too dumb haha. As far as the next test goes, I'd like to see you test the Peon230 Quad by Tech2c (inventor of the hypercube).
Good idea but the Peon230 is actually quite a bad design for 3D printing, almost taking no advantages from the possibilities you have using 3D printing. This is basically a copy of conventionally machined carbon designs with minimal reinforcements at the arms and the base.
Incredible video, this is what I should be doing at work. You might be introdisung stress concentrations at the areas where you change the stiffness of the part, thus a more gradual increase and decrees of the infill might improve thins even more. Maybe someday someone would make a infill that followed the maximum principle stresses. It might be "simple" to combine the Normal stress in X, Y, Z and the Gradient Infill method. (except that you need python the read out the stresses at several x,y,z points. You can du that in ANSYS using the commands, the you could right a command using python run that in post in ANSYS, and get at table of SX,SY,SZ and then read that table with python and create gradient infill. You might need to normalize the stresses though. Also there might be at better way of doing this, since this would be back and forth between sliser, python and ANSYS. ps. I love you work and I aspire to become as good an engineer and home project creator as you, or just 80% would be satisfying.
An approach to the infill problem that I think should have a very high potential and likely should lead to infills that are as close to optimal you will get in a relatively short amount of time (including the fibers you are after), is deep generative learning (a form of machine learning). Perhaps some reinforcement learning-based approach. A quick google search doesn't really yield any significant results, so this seems like an open problem as of now.
An algorithm optimizing the "density of isotropy in the deposited polymer" (avoid the gaps) in the direction of the applied stresses would make sound...
Wow... This was great. Kudos for this video - you constantly keep getting better. This is easily you best video, yet. Both in technical- as also in presentation terms. Keep up with your great work!
Good test about Topology Optimization in 3D printing with 1 axis boundary condition. But I really want to know about how ToOp works in 2 or more axises boundary condition, because you all know about the anisotropic of 3D printing. It will prove a better point about the viable of smart infill. Anyway, thanks for the method you found to add the optimized structure to the slicer. Keep up!
I think the closer the curve interruption is to the complete failure of the part the better. Such a dramatic internal restructuring half way through the load would be inconsistent. You want any changes in the internals of the hook to only occur near the same point it fails, otherwise you'll have an unknown amount of support left internally and the entire hook is unpredictable. And you're missing further possible strength optimization since something is giving out before it could help support the target load.
Very awesome information! I have been playing around with cresting my own infil directly into the design and leaving out infill from my slicer. This is very similar and I'll be playing with it.
In real world use those hooks would have problems with the outer end of the hook breaking off since the reinforcing stops exactly at the midpoint of the curve. Loop ropes through both ends, the hook should rotate a bit, especially when it flexes some, putting the load vector angled toward the free end of the open hook.
Kind of a twist on the multi material idea. What about printing the part with the optimized area as a cavity. Then fill it with something like epoxy. This might even help overcome FDM parts weakest link; layer adhesion.
Very cool project! @CNC Kitchen What I'm missing is information on how much material was used additional on the differen reenforcements you showed. Because if additional material does not matter you could just set the infill to 100%. But a cool result would be sth. like "smart infill gives 20% more stress resistance but uses only 10% more filament (and takes 15% longer to print)" or sth. like that... what do you think? :)
This is an excellent idea! What about trying different materials, for example: a hook printed in pla with the "smart infill"™ printed in petg, so you have a hook with the rigidity of the PLA and the strength of the PETG
Always GREAT vids...but I have a request for a really difficult topic, how to elegantly print nested overlapping structures. ie how did you ultimately get 2 objects to print successfully in this way? Im doing a large drum project and need supporting internal structures to print in this way as well :-)
Brilliant!, This technique would be good for optimization around fasteners, brass inserts, nuts and so on. A little off topic - It would be interesting to check adhesive strengths in the tensile testing machine. I seem to get mixed results on the type of super glue used?
would expect keep the weight the same throughout, or at least under +-5% difference. did not know that fusion had such an optimisation module. have used it only within solidworks.
It would be interesting to run this with an H or P style convergence on the mesh to help increase your mesh density at your high stress areas and open up at less dense areas. Pretty easy to do this in SolidWorks. Not sure about Fusion or Ansys
Would love to see if you could use multi material to get it even stronger by using a reinforced Material( ie carbon fiber or fiberglass, or nylon) for smart infill , this would be a true smart reenforced composite printing
Amazing workflow. I only wish that you had a 100% dense model to compare strength to weight to material cost.
Didn't want to show that hours of adjusting settings could be beaten by yeeting more material out of the nozzle 😆. The weight comparison is nice. But unless it's flying I never cared.
I agree. Strength to weight ratio would be the best way to show the efficiency of this process.
@@mattweger437 Weight = Printing time ;)
You never fail to come up with interesting technical ideas for videos. Well done Stefan.
What an awesome technique! It seems useful for being super efficient with your materials if using an expensive filament or printing a lot of copies of a model, or just for adding a visual detail like the dual color versions you printed. Also, thanks for confirming my intuition that adding perimeters is generally more effective than increasing infill, that little test alone made this worth watching :)
also, it's a great way to show stresses in r&d with multicolor printing.
This is just an accessible version of 'Discontinuity Layout Optimisation' used when designing engineering structures / parts.
He did made a video quite long time ago, that prooves the higher perimeter betten than more infill.
Will you use this in your projets?
On the chance you read through new comments on old videos. A video highlighting whether or not going through that process with the previously strengthened completed hook to see if it will improve it any further.
As a 3D printing hobbyist and a mechanical engineer I can say this is Fxxxing brilliant, the applications are so vast
The computer can run Crysis, but can it run f i n e m e s h s i m u l a t i o n ?
i only simulate my fine mesh at 120fps
What we need is a finite mesh simulation that outputs a weighted mesh that gives the stresses as a proportion of maximum. Then we need slicer integration that will allow you to create gradient infill according to a curve. That way we get better fading between infill strengths so that the load is distributed more evenly and we can fine tune the process at the end instead of restarting each time.
Distributing the load would probably help a ton because it looked like some of the parts began to fail when the internally dense region began separating from the rest of the part. If it could be faded more evenly to distribute load, possibly with custom infill shapes to spread forces into the whole of the part, I bet the part could be made even stronger at the same weight.
Effectively, I think what a large goal should be for the 3D printing community is for us to move away from static infill patterns to infill that is created procedurally for each part to maximize load distribution. I'm imagining a version of infill that has almost that tree-like branching that moves from a thick area near the main stress in a direction that will support the part and then spreading and splitting outward to spread out that load. 3D printing could allow us to, in a way, create parts that have fully connected fiber reinforcement aligned to maximize strength.
I'd imagine optimizing the filament paths is probably harder than just density. Although optimized paths will likely perform notably better than pure density. Especially since strength in various orientations differs non linearly for orientation and density. However, there's already ongoing research into fiber orientation optimization for high end carbon fibre structures... just need some of that for 3d print software.
Spot on. The vertical gradient between 100% infill and 10% infill was a stress riser. Interpolate over some range at the boundary and it'll be stronger.
Building off of this concept, I imagine it would be helpful to have different types of material within the model. PLA and flexible in the same model might provide many advantages.
@Vivid Media Need to complete the loop: AI needs to be trained, AND you would want some sort of feedback loop to improve its performance.
Yay! Always excited for real engineering in 3D printing content!
As someone interested in simulation (via OpenFOAM and SimScale), this was incredibly fascinating! More like this please!
I'm interested
I think I'd like to see more on the shape optimization using multiple materials. In the video you show the one printer printing the hook with the optimized infill in a different color. Instead of another color use a stronger material. So for the base hook body maintain the standard PLA and then for the reinforcements use one of the PLA filaments that has adders in it, or print that section with say ABS or Nylon. I am not sure how the two material bonding would react but it would allow you to use a cheaper material for the base part and then a stronger but pricier material for the areas that need the reinforcements. Similar to some injected molded parts where there are metal reinforcements in critical sections of the model (one thing that comes to mind is metal reinforced feed lips on polymer AKM magazines).
Great work as always sir, from one engineer to the other, I love the way you structure these videos and tests.
Good idea, but rather than different plastics which print at drastically different temps I would suggest reinforced vs non-reinforced plastics. I.e. use standard cheap pla for most of the part, but on the high-stress regions use carbon fiber/ Kevlar/ cellulose reinforced pla.
@eugenioLU91 I haven't seen that, however I question whether you're thinking of structural or decorative pla. Either way it's a better place to start than attempting to print abs into pla (which will melt straight through the surrounding pla producing little more than plastic blobs).
@@xxportalxx. PET-G and other copolymers can print at about the same temp (250) as CF-nylon so this might be an interesting combo.
@@robertasumendi I haven't had opportunity to use much of either, however the ability to mix materials seems like a promising niche for additive manufacturing.
Personally I've been spending more time researching the potential of printing protoboards easier (as an ee that is more aligned with my use case)
Also adding in flexible filament and seeing the affects of that.
It's obvious the reason why the "full optimized" design was strongest is because it delaminated before it could fail in tension. That allowed the structure to deform and relieve some stress.
Very interesting. I cant help but wonder how this compares to 100% fill of the whole hook. Also, then reduce the thickness or width of the hook to get back to the same weight at 100% fill.
You are my gurú! i work in a 3D printing laboratory here in México ,so whenever you come to Yucatan you have a room reserved in my home! you inspire us buddy!
Hey Stefan. From my understanding, the main factor that you want to keep in mind is the contact surface area in the plane perpendicular to the force vector. Hence you want to optimise force/(surface area). Solid infill certainly increases the contact surface area. The other factor that you might want to consider is the print direction for the infill in those place, as you have previously shown that it matters a lot (i.e. layer adhesion < material strength).
What an interesting group of thought/structural experiments. I'll have to keep those tricks in mind.
One thing I've done for parts that I need to tap (When we make samples of castings at work) I cut a bunch of .02mm lines radially around the hole so that the printer does a zigzag pattern around the hole to be tapped. This way I can force plenty of solid plastic around the hole to finish drill and tap without having to up the perimeters everywhere else.
Great, this is basically using the "virtual fiber" technique!
The displacement sharp edge you found is because something called triaxial state of stress.
The hooks are failing like simple engeniering materials because the rupture is always gonna start in the inside (like a relatively ductile material). It happens because there are not so much stress concentrators (like the sharp edges of the infill) and the deformation is not possible because the surrounding infill is dense, blocking it.
I'm sorry because the broken English. You can find more research material on this searching the subject fracture mechanics.
Danke für die interessante Ingenieurssicht, ohne Staubfänger ;-)
Wow, this video is amazing! I really think this kind of work unlocks the potential of 3D printing to use material effectively.
Slingshot might be a good demonstration object. Interesting shape, several load and structure positions, exciting results when they fail.
As someone familiar with the use of FEA this is an interesting topic and workflow. It would be interesting to see how the parts reacted when subjected to loads that it wasn't optimized for compared to the standard infill parts. At least with a hook it's fairly easy to know how the part will be loaded, but sometimes parts can be subjected to loads that you may not suspect, hence the need for safety factors. However, I'm sure this consideration is probably overkill when you're worried about hobby projects, but something to keep in mind nonetheless.
Nice. I like that the reinforced ones show distortion some time before failure. Sounds like a good safety feature to me.
Quadcopter arm optimization would be awesome
I was thinking exactly the same thing!
Yeah, but that would be really hard. The case where quadrocopters break usually is in a crash, so in my opinion you would have to optimize for this case. Using this method with FEM simulation, you only optimize for ONE case, usually where the force is one way and constant... Also in flight, the forces vary, if you change direction, because of inertia (Trägheit). So you would have to combine optimized infills of many cases. That might be interesting.
Very interesting solution and very impressive results! This would be an important add-on plug-in for slicers.
Also works on Ondsel/Freecad; the tool is called Scalar clip filter
if mesh is not broken
Design 2 parts in the sketch model and set infill for each separate part then combine them before slicing is a quick workaround and less time. once you see the warning infusion design a simple part to cover that area and import both models then set different infill for each presto a simple way to do it without the complication I've done this for years on mech parts.
Stefan - the one nice thing about the partial fail before full fail is if you had this in an industrial situation - you could in theory monitor and take action for that 'warning sign' before deciding the part was too close to it's tolerance limit before actual failure. On the other hand, it means you have to leave a larger 'reserve' in strength vs. a 100% infilled part which may make the savings (weight, cost, performance) trade off not worth it.
To create acutual fibers, you could generate a 3d grid of fibers and then use the combine/intersect tool to just get the parts of the grid, that intersect the simulation mesh. Then you could use this in other slicers, as it is baked into the model.
Ive done stuff like this in the past with Fusion. Never tested like you did but you did a great job explaining the process along the way. As for the gaps on concentric infill, what about using a larger nozzle diameter and low layer heights? This will effectively reduce any voids in the part. Also in fusion you can modify the shape optimization criteria for a higher target mass so you can get an optimization that is all connected. Which will likely reduce voids going from solid reinforcement to infilled spaces.
A very simple way of adding the smart infill :). Great work!!! An idea I came up with, but lack of time has stopped me :(. I was looking into using a basic fractal branching set of connected rods like bone. So the sides can spread the stress more evenly. I also think it’d help if the part was PLA and baked. As to what part you should try, I vote for the quad copter ;)
You're a smart young man Stefan. I'm new to 3d, just having gotten my first printer last week, but have been watching your shows for quite some time. I Love your content man!!!
This is one of the best things I've seen in 3d printing in the last few years.
Wow, imagine implementing this on 3d printed metals to make parts lighter but still as strong as a solid part(or stronger since there's room for strain relief within the structure)
There is a number of industrial software available in this regard. The keyword here is topology optimization. But sadly this really is not for the DIY community with prices starting from 10 k to well over 50 k € per year.
withinlab.com/software/new_index.php
www.ntopology.com
www.materialise.com/en/software/3-matic
... the list goes on
This is genius-level. So good! Can’t wait to try this out.
16:40 Das was du meinst ist ein schön zu sehender ausgeprägter Bereich nach der Streckgrenze (siehe Lüdersdehnung). Das tritt vor allem bei zähen Kunststoffen (und niedrig legierten Stählen) auf. Dieses Verhalten ist bei zähen Kunststoffen stark von der Temperatur und Belastungszeit aber auch durch Kerbwirkungen in der Typologie abhängig. Bis zu diesem Punkt der Belastung ist die Dehnung komplett elastisch, es tritt also nach wegfallen der Last bis zu diesem Punkt keine plastische Verformung auf.
At 13:40 you can see that the hole in the bottom left that was supposed to be normal support infill is entirely 100%, though an isolated hole in the center in the modeling was not intended to be so.
This is outstanding! GREAT work. Thanks for sharing. I've worked with topology optimization for my master degree, studying the application of top-opt techniques in the early phases of an industrial design project (how to deal with style, ergonomy and so on…). This is such a clever application!
This is really interesting. We've been using Slic3r's volume-based variable infill to solve issues with part strength so I'll give this a go and test the relative differences in strength. Thanks!
Great, let us know your results.
since forces here are somewhat in one direction, infill could be done in that direction as well, in all cases (vertical). ie. horizontal density 20lpi, vertical 7 lpi. so that more vertical strands are present. also, continuous strands have more strength so they could be as solid as possible from top to botton, in an S, J or whatever shape needed.
Great Idea! I used Simplify3D and 2 combined "layer by layer" print processes (The main model with a sparse infill and the optimized meshbody .obj at 100%) with a similar result.
In S3D one can just import the "optimized" .obj file as a "modifier" file.
You could adjust the Fill Angle infill setting to orient the reinforcements rather than using Concentric.
Impressive work you have done
Thanks for sharing👍😀
That's the most interesting addition I've seen since the last few years in 3d printing. It would be really nice if this feature would be added directly to the slicing softwares. It would be nice to see the increase of streingth of a 3d printed object combining the effect of optimized infill and copper electroplating.
What you observed in your final graph is not a release of internal stress as you hypothesis. In engineering we call it the elastic and plastic stress strain curve.
The initial part is elastic until you reach the drop meaning it will be permanently deformed. The plastic is just the state of permanent deforestation until breaking.
This looks super useful. Think I might give it a shot and see if I can use it in future prints.
This is definitely one of the great advantages 3D printing can offer, excellent video!
I know this is a stretch but I wonder if it is possible to use the pathing and infill data from the slicer as real geometries inside of the simulation tool (as internal cylinders perhaps?) to produce much more realistic simulation models.
I liked the video the moment I saw the display picture 👌😄
That was a significant strength increase, fascinating topic.
Also a great way to optimize. Most strength for minimum material.
What I'm curious to know is what would the strength test be if you made the entire part with 100 percent infill. Feel like this would give you an upper limit on what's possible in terms of strength with the given material, and provide you with an idea of how close you can get to a theoretical maximum strength.
The hooks that had two bumps on the end of the load curve exhibited what is known as lack of composite action; i.e., the inner and outer structures did not act as a unified assembly. It looked like the outer skin failed under flexure followed by the in-fill core failing under buckling or shear.
It would be interesting to see how the smart infill pieces strength compares with itself, after annealing them. Great video and awesome technique, hope to see (and test) more of this!
It is great to see that topology optimization is finally becoming popular (it exists since 90ies). Great application! Some comments: Ansys has also topology optimization integrated (it had it already 10 years ago, but it seems they fogot it)
There is also topology algorithms to optimize the orientation of fibers like CAIO, this might be usefull to align the printing orinetation. Unfortunately the software developers are a little bit slow to implement it.
There is also latice structure optimization. As far as I know only altair offers software for this and it probaly can't concidder the anisotropy of fdm.
You really bring value to the 3d print community. This was super interesting, i'm going to test it
Wow, Ansys :D Its like dropping a nuke on a plate when you want to pop some corn, but I liked the video :) Keep up the good work!
One of the best 3dprinting videos I've ever seen.
Nice vid, a 100% infill would of been good to test as a comparison.
I second that.
Neat! I already do that with my prints of I know there is a point which will need reinforcement.
I was just working on a functional print and this came up, which is exactly what I needed since it keeps breaking at a certain area
Boom! This is so great. We were just talking about this in the studio yesterday. So awesome. Love the style and speed of this video! Great work!
This is really cool !
Exporting the optimized mesh and using modifiers is quite a nice workflow
I think making sure the modifier has a good connection to top/bottom/perimeters would be in of high importance. Having the modifier connected by standard infill to the shell will probably have lower performance. Perhaps using the shape optimization mesh as a guide to hand model the modifier would yield the best results.
Those results were both enlightening and impressive. Thanks for sharing and keep up the good work.
This is an amazing video!!! I really appreciate the work you into it and it's a testament to how proper engineering techniques can yield great results.
At a bare minimum, any part I print that requires strength will likely be the 5 perimeters/10% infill, which seems to be a nice labor/strength improvement tradeoff.
Hallo Stefan!
Als Bauingenier kann ich dir sagen dass du wirklich tolle Arbeit machst. Eines möchte ich dir aber bezüglich der "mesh-size" mitgeben: Spannungsspitzen können zum Teil sehr hoch sein und auch dort auftauchen, wo dein Netz gerade keinen Knoten aufweist. Im Bauwesen ist das meist bei Ausnehmungen , Einsprünge und dort, wo Stützen sind. Hier würde ich das Netz etwas kleiner ansetzen um wirklich die Spannungsspitzen zu bekommen und zu handeln.
Was auch eine Rolle Spielen kann: das netz geht von einem Festpunkt aus und kann daher andere Ergebnisse bieten, je nachdem wie das Bauteil verschoben wird, da andere Knoten berechnet werden.
In wiefern das beim 3d-Druck mit anderen Problemen (layeradhesion usw) eine Rolle Spielt lass ich mal dahin gestellt ;).
Lg Thomas
Great channel, great video. At last someone who does real technical evaluation rather than just a personal opinion. Respect to a real maker.
This video was absolutely fascinating... You could use this method for a lot of applications. I got a little confused towards the end of the results, maybe you rushed through them too fast, or I'm too dumb haha.
As far as the next test goes, I'd like to see you test the Peon230 Quad by Tech2c (inventor of the hypercube).
Good idea but the Peon230 is actually quite a bad design for 3D printing, almost taking no advantages from the possibilities you have using 3D printing. This is basically a copy of conventionally machined carbon designs with minimal reinforcements at the arms and the base.
what you need to test is a gradual change from 100% to 10% to stop fractures at boundaries which is where it seemed to fail
Incredible video, this is what I should be doing at work.
You might be introdisung stress concentrations at the areas where you change the stiffness of the part, thus a more gradual increase and decrees of the infill might improve thins even more.
Maybe someday someone would make a infill that followed the maximum principle stresses.
It might be "simple" to combine the Normal stress in X, Y, Z and the Gradient Infill method. (except that you need python the read out the stresses at several x,y,z points. You can du that in ANSYS using the commands, the you could right a command using python run that in post in ANSYS, and get at table of SX,SY,SZ and then read that table with python and create gradient infill. You might need to normalize the stresses though. Also there might be at better way of doing this, since this would be back and forth between sliser, python and ANSYS.
ps. I love you work and I aspire to become as good an engineer and home project creator as you, or just 80% would be satisfying.
What about simplify 3d how do I isolate and print the high density modified mesh
An approach to the infill problem that I think should have a very high potential and likely should lead to infills that are as close to optimal you will get in a relatively short amount of time (including the fibers you are after), is deep generative learning (a form of machine learning). Perhaps some reinforcement learning-based approach. A quick google search doesn't really yield any significant results, so this seems like an open problem as of now.
Nice video, good to be back at school sometimes and learn really interesting things! Thanks!
Oh man, I love this stuff! I've already learned so much useful information from your channel. Thanks for all you do! 🙂
Interesting hack of the FEA!
An algorithm optimizing the "density of isotropy in the deposited polymer" (avoid the gaps) in the direction of the applied stresses would make sound...
Wow... This was great. Kudos for this video - you constantly keep getting better. This is easily you best video, yet. Both in technical- as also in presentation terms. Keep up with your great work!
You just blew my mind. A brilliant idea. All I can say is thank you.
Good test about Topology Optimization in 3D printing with 1 axis boundary condition. But I really want to know about how ToOp works in 2 or more axises boundary condition, because you all know about the anisotropic of 3D printing. It will prove a better point about the viable of smart infill.
Anyway, thanks for the method you found to add the optimized structure to the slicer. Keep up!
Wow! This is an amazing idea -and I you can even do it in Cura.
How can you make the stl piece and the obj piece match? When I import them on cura I see them as two separate pieces
If your modifier is using multiple perimeters (3-5), you'd likely get better results than pure 2 perimeters and 100% rectilinear.
The specific strength my man, the specific strength.
I think the closer the curve interruption is to the complete failure of the part the better. Such a dramatic internal restructuring half way through the load would be inconsistent. You want any changes in the internals of the hook to only occur near the same point it fails, otherwise you'll have an unknown amount of support left internally and the entire hook is unpredictable. And you're missing further possible strength optimization since something is giving out before it could help support the target load.
Very awesome information! I have been playing around with cresting my own infil directly into the design and leaving out infill from my slicer.
This is very similar and I'll be playing with it.
In real world use those hooks would have problems with the outer end of the hook breaking off since the reinforcing stops exactly at the midpoint of the curve. Loop ropes through both ends, the hook should rotate a bit, especially when it flexes some, putting the load vector angled toward the free end of the open hook.
Kind of a twist on the multi material idea. What about printing the part with the optimized area as a cavity. Then fill it with something like epoxy. This might even help overcome FDM parts weakest link; layer adhesion.
Very cool project! @CNC Kitchen
What I'm missing is information on how much material was used additional on the differen reenforcements you showed. Because if additional material does not matter you could just set the infill to 100%. But a cool result would be sth. like "smart infill gives 20% more stress resistance but uses only 10% more filament (and takes 15% longer to print)" or sth. like that... what do you think? :)
That's one hell of an engineering project, Stefan. Brilliant job!
This is an excellent idea! What about trying different materials, for example: a hook printed in pla with the "smart infill"™ printed in petg, so you have a hook with the rigidity of the PLA and the strength of the PETG
You are awesome Stefan! This bloody useful for me, because I do a ton of practical 3d printing.
Always GREAT vids...but I have a request for a really difficult topic, how to elegantly print nested overlapping structures. ie how did you ultimately get 2 objects to print successfully in this way? Im doing a large drum project and need supporting internal structures to print in this way as well :-)
Amazong work Stefan, I love learning this type of stuff from your channel. Just superb! Thank you!
Brilliant!, This technique would be good for optimization around fasteners,
brass inserts, nuts and so on. A little off topic - It would be interesting
to check adhesive strengths in the tensile testing machine. I seem to get
mixed results on the type of super glue used?
Check this out: ua-cam.com/video/JSff8OMRHtw/v-deo.html
Seriously fucking excellent work, Stefan. You rock and embody the reason why Germans are awesome.
watching this again and now i am curious about using this with multi material to conserve expensive exotic filaments
This is SOOO COOL !!!! Thank you so much for teaching us how to do this!
YOU ARE AMAZING!
Excellent video Stefan. I love videos that mesh 3D printing and stress analysis. Thank you.
Wow, this is really useful. Gonna try this asap for some of my mechanical designs.
Note that most topological optimizers evaluate stiffness, not strength.
This video was incredible. Probably the one I've enjoyed the most. Thanks
would expect keep the weight the same throughout, or at least under +-5% difference. did not know that fusion had such an optimisation module. have used it only within solidworks.
It would be interesting to run this with an H or P style convergence on the mesh to help increase your mesh density at your high stress areas and open up at less dense areas. Pretty easy to do this in SolidWorks. Not sure about Fusion or Ansys
Would love to see if you could use multi material to get it even stronger by using a reinforced Material( ie carbon fiber or fiberglass, or nylon) for smart infill , this would be a true smart reenforced composite printing
Amazing! :o
Thanks from all the 3d-printing architecture students!