Thursday, 23 March 2017

3D printing introduction (Anton)

Introduction

My printer quickly is FDM (Fused deposition modeling) type, means that it material (filament) is pushed with a extruder motor into the hot-end where it melts and then pushes  through small hole inside the nozzle onto the hotbed. There are dozen of other types of printers but FDM are the cheapest and well accessible.




It's sub 300 euro Prusa i3  mk2 clone made by hictop as a DIY kit, I had some conditions as minimal use of acrylic part (because they are brittle and split often), solid metal frame (so it will be rigid), heated build platform so I can print more demanding filaments in the future, support for multiple print heads but have only 1 stock so the head is not heavy. Heated bed be supported on 2 rods instead of one, I didn't wanted to have issues of bed balance when it's not fastened on multiple corners and might bend a bit. Problem is that for good first layer adherence which is very important a bed level accuracy needs to be down to 0.1mm  and I can't see that happening if the bed is supported only by 1 rod and you can bend the platform by 3mm only with your finger. So I'm very happy these conditions can be meet under 300 euro and because I omitted ready-to-print instead of saving on materials this kit saves on labor. Meaning that other clones use cheap threaded rods for Z axis while this kit uses lead screws which are far superior and do not have issues like backlash. 

Baby lamp

As first attempt I tried to print a break on every single joint "snap together" mini lamp to evaluate how easy difficult it would be to make a baby lamp as part of the meerkat family. I used flowing thingiverse files:





On the pictures it looked lovely and easy, but my experience was far away from expected. First I had handful failed prints:


And then in the end after couple days of frustrations even with finished product I was not very happy:
It's not possible to move this lamp freely, and in essence every single joint failed. Bigger snap joints broke, smaller snap joints deformed permanently. The print could slightly bigger margins between the joints and my printer could be better as well. But most importantly my filament was not suited. Was extra disappointed because the author of the file is genuine Makerbot had no proper instructions how to print it "Add light. Smile" are joke and not a instructions, what about recommended filament type, its temperatures and print speeds, recommended layer height. Print speeds may wary between the printers and temperatures warty between the filaments, but at least having 1 detailed specifics which gain good result as a reference would be helpful. Sadly this is reoccurring theme in the 3D printing community, a lot of time coolness goes before practicality. Which made me investigate more about material properties.

Materials and stress strain curve

Before to me a hard material was hard stiff strong material which was opposite to soft elastic one. But there is so much more to it, the material properties are often correlate together, but that doesn't mean they are the same. For example for me hardness mean actually toughness and I had lot of terms mixed up and probably even now still sometimes I do. Even in this article I could have make mistakes, I investigated this topic for about 5-6 hours, I have only surface understanding so I apologies to people who know better and see that I talk nonsense.

Toughness 

Toughness is measurement how material absorbs energy on impact, it's opposite to materials being fragile. The difference when you drop your lunch from your workbench (totally not speaking from my own experience). The ceramic place will shatter even when it's harder, while the fork will only bend a bit.

Hardness 

Hardness is the ability to resist penetration (imagine trying to saw a glass sheet, the saw will almost glide, but not cut). If you watch some metal working YT videos you see the artists to heat meat and then rapidly cool it, often they use files to test the hardness and you can see the file glide instead of "bitting" into the metal.

Stiffness

Stiffness is telling how the material will deform when stress forces are applied. Often stiff materials are hard, but it doesn't have to be, and that's why I started to mix them up between each other. Often they are more fragile which means that they can shatter much easily. That means that they will not deform under the stress but might fail catastrophically.  And interesting material characteristic is the the strain curve, the slope and shape of the curve describes how stiff and tough the materials are.


This is a simplified stress strain curve:


When we look at first portion of the plot we see how stiff the material is, any deformation before the yield point will return to original shape. If the yield strength is applied the deformation will not return and deformation will stay. Reaching the ultimate strength means reaching the material's critical catastrophic failure point. Some materials can be stiffer and having sharp slope in the first part of the curve, so they will not deform under the stress. But the line can be shorter so they will be able only accept smaller stress forces before the yield point. In combination with short curve after the yield point now the material might be stiff but very fragile and probably undesired. The will not move / deform elastically they will not move deform permanently much and they will fail will little warning.  This is the reason glass is not used as structural material. And this is the reason why after hardening the tempering is often done. Hardening when the material is heated and then rapidly cooled extends the yield point further but decreases the curve after the yield point making it very brittle. The hardening creates internal stresses in the material because when materials heats and cools it expands and contracts. Doing this rapidly will coll the the outer layer faster then inner layers so they will be under different dimensions as the colled down layers, creating inner pulling forces which changes the properties of the material. Then often tempering is used to relieve bit of this stress which is heating the material again but leaving it too cool down much slowly. Final effect is something which is harder as original and not as brittle as the hardened-only material.

In the end the brittle property is so undesired that even the yield point is sacrificed just to have something more tough. Brittle materials are no no for many applications and double no no for structural support.

Very interesting videos are online under topics of hardening / tempering and strain curves as well. Interesting video is by Destin showing prince rupert's drop which might look as unbreakable but when it gives it gives catastrophically releasing all the inner forces.




Why is this relevant?

Because I realized that I have strange filament supplier, I enjoy his filaments, but his PLA behaves a bit different than expected. This is all relative (and comparing it against metal is no point, it's more referenced between the plastics). PLA has to be this weak material which can't stand temperatures and should be weaker than ABS and yet my PLA prints can hold bigger abuse than my ABS prints (maybe my ABS supplier is very bad). And even the temperature sensitiveness is bit strange, I replaced my filament feeder made from ABS with PLA part and it holds even better, touching physically the hot-end which goes close to 200C. Or filament cooling duct which is few mm above heated bed which is around 70C. I was expecting they would bend, deform and yet after many kilograms of filament they still hold pretty well. 

But what is the negative and main reason of this blog is that the filament is much stiffer and brittle than expected. Meaning parts which are meant to snap together will not budge or deform temporary for a moment, they will need large force / stress and still instead of bending will just snap and break permanently.

This is important to know when designing parts that the material might break when not expected, which doesn't make it ideal for structural parts I wanted to do with it and frankly put me off a bit. First initial idea was to make whole lamp as 3d printed object which after these experiments and realization how brittle the parts are is now canceled. And even for smaller parts I'm still bit unease, because I studied how much effort people put to avoid brittle materials in structural parts and yet here I'm doing the opposite.

Is there a way to predict yield points? 

Alternative title: My elaborate excuse why you shouldn't blame if the parts will break and why you should be extra careful around them.
This would a bit mixed answer, Yes with a hint of no, give abused elements more material (thinner structural peaces can't hold as much as thicker). But still it might be upto the way it's printed, the objects have weakness in the layers, like grains in the wood the 3D prints can whitstand much bigger forces from different axis than going againts their layers / grains. Often shape will dictate which way the model should be printed (often the larger footprint first) because 3D printers have difficulties to print onto thin air, it's more desired to have solid layer below so printing pyramids upside down is much harder than the right way up. Mechanical requirements might dictate the design, the design shape might dictate the way it should be printed and that might then go back and dictate the mechanical requirements because some stressed part might be printed in a very weak way. So it's kind of infinite circle without end and even changing orientation might solve given weak points just to create complete new weak points. Parts will break more likely on the Z axis than on X or Y. I can predict which parts are the weak point (often the whole part) but it's harder to predict how much force it will take, it changes depending on many aspects. But overall splitting two layers apart is much easier then breaking all layers at the same time.



There are ways to go around but they don't really work. Increase the layer size (bigger less fragile layers make stronger print), but I already print at the highest recommended layer size for my nozzle type. Use of acetone bath to melt the layers a bit to allow smoother surface and stronger bond. But acetone works only with ABS while with PLA it has very strange effects which I can't call desired, there is limited use of acetone with PLA. Baking PLA will make it ultra strong, but there is chance of changing dimensions of 10% which is unacceptable for mechanical parts and there is last resort which is using 100% infill, first it increases chances of failed print, because my printer doesn't like to print so dense and is extra slow (printers are slow by general but mine is even slower than that). Increases print times and consumed material as well and yet they will not increase the Z as much as X Y. So everything is compromise, the the mechanical parts might not be as precise (quality) or they might not be as strong, but anyway it's still weaker than other fabrication processes. Having CNC machine and solid block of aluminium and we wouldn't have such issues.And they would be much easier predictable, inside the software.

Software like SolidWorks is like a dream for engineer, it supports large spectrum of tools SolidWorks can predict lot of problematic places from stress viewpoint, from ease of manufacturing viewpoint as well. If the work is made on a lathe it can be analysed if it has some problematic places, or if the work / part will be molded if there are parts which can damage the mold or make removing the part problematic. If part is printed with FDM aproach then it will hate overhangs for example and it's lovely to see these problems before even trying to print the models.

Finite Element Analysis is very cool as well:

But is absolutely not supported for FDM filaments, because there is no reliable and sensible way to predict how it will behave when the objects are not really solids but glued / laminated layers together. This is why MDF, genuine wood and ply wood behave differently. For example ply wood will be weaker on one axis as well. This is harder to model but what it complicates it most that for all other materials the solid block of work is solid, with 3D printing a solid box might be printed almost hollow, so after designing the part lot of the properties change in the printing process which SolidWorks can't forsee. Multiply it by different filament types and by different suppliers which can change between the badges as well. So there is lot of variables which change a lot, even bad extruder calibration or uneven Z steppers might not show visual difference but might cause under-extrusion in some parts and cause unnecessary weak points. Bit worn down nozzle might mean that the layers will bit bit more thicker but not as tall as expected meaning the Z layers will be even weaker. There is too many variables, too many unknowns and many properties changed AFTER the analysis to do any proper predictions and this might be one of the reasons why there is no FEA on FDM.

So approaching it with attitude "It's very fragile and brittle. In past all fragile materials were avoided for structural support and we are using it anyway. So it might break at any time" is probably the best, trying to quantify it can lead to misleading security. Other problem of reinforcing a sections of the part will just make some other section the new weak link, and everything is compromise between many factors, so the easiest approach would be not to abuse the parts at all.





No comments:

Post a Comment