Thin Walls

Bill,

A part of the wall thickness issue is simply a constraint of the process and the machine.

A printing head has to deposit material and this material must be surrounded by support material. The size of the drops, laser, extrusion nozzle etc. all have a minimum thickness. Any thinner and the structure collapses, the wall has holes etc.

These wall thicknesses are hard limits of the process itself. Over time they will improve.

But, as people have stated there is a difference between walls and details.

Where we made errors is in becoming too stringent in applying these and higher standards to adventurous models.

A wall is any structural element of your model.

So lets say you're printing a porcupine. The spines of the porcupine would have to be 0.7mm thick at least(in White, Strong & Flexible). For such a structure I would even recommend 2mm. Any thinner than 0.7mm and the pins simply would not print properly. Thicker than 0.7mm they should print but might be fragile. 1mm would be stronger, 2mm stronger still.

But...the spines each have a very fine point at the end of them. This is a detail and could be finer and thinner than 0.7mm. Because this detail is not a structural element the printer will simply try its best and then leave a tapered off end. This is similar to what happens with corners, edges etc.

So lets look at Laurana: https://www.shapeways.com/model/7842/laurana_50k_cropped___sh elled.html

The entire structure is a wall. This simply could not be any thinner or it would fail. The eyebrow is thinner at the very end, this is a detail.

Anything that if removed would cause a hole to appear in your model or mesh is definitely a wall.

Anything that can be safely removed in and of itself without creating a hole in the mesh is most probably a detail.

Details taller than 1mm, are walls.

Joris

Sorry, but in a passed life I used to help design ships, so I have a basic understanding of structural engineering, though I have forgotten all the maths.

Now the interesting thing about a large ship is that the plating on the outside is vary, very, thin compared with the volume that it encloses, but that there is a system of structural members inside the plating that supplies the structural integrity.

So applying this you our models, we could build a lattice out of robust members without any problem. If we then added a thin sheet to one side so we had a series of open boxes we could still print it if the sheet was thick enough. The problem is that there is nothing that I have read that gives me a clue about how thin I could make this sheet. I know this will be complicated because the further apart the lattice members are the thicker the sheet needs to be. Also the span of any particular thickness that could be supported would depend on the size of the lattice members as well as the distance between them. Would I be right in thinking that the thinnest possible dimension of this sheet would be close to the minimum feature size?

Software that only measures arbitrary minimum thicknesses will always give false negative in cases like our lattice, unless structural integrity is taken into account.

Bill,

I seem to not be explaining myself correctly.

I'm sorry for being confusing.

The thinnest the sheet could be is the minimum wall thickness. The thinnest the lattice could be is the minimum wall thickness.

The printer in principle can not make anything thinner than the minimum wall thickness.

However a detail such as a sharp point can be produced because the printer simply gives up at one point leaving some of the point intact.

An edge of an eyebrow could remain but if a wall would have the same thin features it could not be produced.

Sorry for using words like 'structural element.'

What I'm trying to say is if you're trying to stack golf balls one on top of another your structure would at minimum be one golf ball in width. You can not stack half a golf ball. This golf ball width is the minimum wall thickness.

And yes, if you look closely at some points on the stack of golf balls you can see that the width is less than one golf ball. But, if you'd want to add a layer to your golf ball stack you can not add half a golf ball. You can only add one golf ball.

Joris

Now I really don't understand what is going on as I have a lot of white detail pieces where the minimum size is about 0.6 x 1.1mm.

Unless, of course the minimum wall for white detail is really around 0.5 or 0.6.

I'm still somewhat confused about the "structural" definition. Is the goal to make the object structurally sound in the "will not fall apart under its own weight and/or shipping" sense, or is the issue "will it survive the compressed air post processing"?

For example, the three objects below:



Assume the blue material is 1mm x 1mm in cross section, and we are using WSF. The object on the left should be acceptable, as no subcomponent is less than 1mm.

The structure in the center adds a red piece which is .5mm x .5mm. Adding this piece only increases the overall structural strength, but the red piece - by itself - is small and potentially delicate.

The structure on the right has the red piece "exposed" and not braced against other components.

Which of these would be printable?

Today I received another "walls to thin" hint.
The difference was, that there was a photo added which points at the troubled places.
In this case I made the error since the object was only designed fro white and flexible, but to thin for others.
This leads me to the proposal of storing the minimal wall size of an object and restricting the materials for printing:
An element < 0,7 has no materials
elements between 0,7 and 1 only white and flexible ....
elemnents between 1 and 2 only ....

That would save a lot of troubles. Is that considerable?

Woody64

I've got time to work on 'em now, so I've started replacing my Not For Sale models in my shop. Two down, three to go...

I'd appreciate it if you'd glance over this one:
https://www.shapeways.com/model/56589/jovian_schild_class_esc ort_cruiser.html

before I work it over, since it shares a lot of geometry with this older ship here:
https://www.shapeways.com/model/30103/jovian_lanze_class_stri ke_carrier.html

That one has been successfully printed very recently.

While I greatly appreciate that what you guys are telling us, and your efforts toward educating us to build better drawings, I would have to fully agree that there should be a way for the customer to request that the model be printed anyway. With such a request, you give up your rights to a reprint, unless you pay for another one.

The other thing that I've struggled with is that my most recent experiences have been with a model that has already been printed successfully and was then later rejected on a subsequent printing request.

I realize that the process is evolving, but this was a big surprise and now instead of sending out product I'm back at the drawing board.

Again, you guys (the fantastic Shapeways team, Joris and Bart rock!) are doing a great job helping us all figure this out, but sometimes we just need to get something done.

Does anybody know an actual status?

Is there now a check done after/when uploading?

Do we get a hint where the problems are.

Is there a correct description of "detail" and "wall" now.

Are the old models checked, to prevent that another customer buys an item and get the response walls too thin.

If not. Is it guaranteed that the designer and the customer come into contact to check corrections possibilities?

Woody64

Seems to me like there should be a simple solution. If you want your piece printed "Best Effort" then it goes into a pile of similarly flagged pieces; once there's enough Best Effort pieces for a full run, do one with just those.

This means that if you flag Best Effort you are accepting not only that you might have made a mistake, but there's a chance that someone else might have designed a piece which breaks the design specs and ends up messing up yours. [I don't know how likely this is; if Joris or someone wanted to weigh in with that data, that'd be helpful. It seems unlikely, at least in the SLS method.]

I'm guessing that there would in fact be enough of those that the Shapeways team would end up running one or more such batches per day - I don't know for sure. Presumably, if this is not the case, then Best Effort may also mean that you have to wait until a full batch is going to run, so the usual ten-day guarantee won't apply either. Wouldn't expect it to be a big deal, though.

So - add a checkbox for "Best Effort" to the upload page (with a "What's This?" popup explaining the terms). Checking this means that if your design uploaded successfully, it will go to print - when there's time and enough other Best Effort pieces stacked up. And you'll receive the result, whatever the result is. But it will be at risk from more than your own errors; that's the price you pay for having it bypass the checking process.

Does the new wall checking software find all instances of thin wall or is it stop after the first few hove been found?

Bill - unless I'm mistaken, there is no automatic wall checking software. The feedback I've received has always looked like someone manually looked at the design and identified the thinnest parts, and then measured them.

As it happens, though, I'm 90% done a Blender script which will in fact automatically check wall thicknesses... which is then 50% of a second script I have planned, to automatically fix them too. It's slightly approximate, but should be much better than visual inspection (and faster too). I'll be posting a link to the script in these forums as soon as it's ready, I promise - should be within a few days, could even be today.

Great! More arbitraryness.

Just wanted to suggest that we give SW a bit of a break, here.

The "arbitrariness" should not be entirely (or even mostly) blamed on Shapeways. Wall thickness causes problems because of a) the laws of physics, b) the realities of handling during the printing process, and c) the capabilities of present-day printers.

a) Physics. You can easily model something that can't support its own weight. If you think that a manual check is arbitrary, please suggest any non-arbitrary way to predict -- without actually printing your object first (too expensive!) -- that it will or won't collapse even if left on its own. Humans are imperfect, but can assess whether objects are likely to collapse, bend, or chip better than a script. Unless you know of a script I'm missing.

b) Handling. Let's say you've written a decent finite-element analysis tool that can determine whether an object will support its own weight. Now: does it properly simulate the stresses when an object is lifted out of the printer, packaged, and shipped by UPS? Does the metal-print simulation properly simulate the un-infiltrated "green" state and the pinch of tongs reaching around the object? Etc. etc. Here, a well-informed human is your best bet.

c) Printers. Very tiny features can cause problems which can disrupt an entire print run. The basis of SW's business -- the reason their prices are so low -- is the ability to print many people's objects in one run. Printer-related model problems are probably most amenable to automated script detection.

I'm sure SW is deeply interested in reducing the arbitrary nature of some of the checks. Perhaps a fuller appreciation of the extraordinary difficulty of automating these checks would improve our patience and happiness. This process is really the difference between modeling stuff in your computer (almost anything is possible) and making it real (most things you can design are not printable).

Reality is arbitrary. The checks are primarily reality checks. There is no such thing as an automated reality check, and there won't be for a good long time.

The accusation of aribtraryness comes from having bags full of pieces that have been printed but now fall foul of the minimum thickness rules. Almost all the professional suppliers (casters, photo-etchers etc) I have deal with supply me with pieces that have failed so at least I know what when wrong and I can design around the problem.

One of my Snowflake balls was rejected for thin walls, apparently because the text on the bars was too thin. When I pointed out that the text should print as part of the bar, adding its thickness to that of the bar, I was told that the text was "loose" and would fall off. The text was done as a separate mesh (or meshes), but it is immediately adjacent to the bar, so I would expect them to fuse together. Am I misunderstanding how the technology works?

Another Snowflake ball, ordered by Robert Schouwenburg and done exactly the same way (but with different text), is now in production, having apparently passed the printing checks.

I think you have just found one of the universal laws of unintended consequences. If you want to make two part that intersect but do not fuse you have to have a bigger gap that you would really like, but if you really want the two part to fuse they will always separate.
In general, if you want to build something without moving parts it should be a single mesh.

Hi GHP,

I think you just need to not build models that rely on fusing of adjacent parts. If you want two meshes to print as a single object, you must explicitly overlap them in space.

That is, even though two parts are 0.3mm apart, which will cause them to print as a fused object due to the statistics of the printer's resolution, SW will not allow you to claim that this printer-error-induced connection constitutes a real structure. (I hope you would agree -- what if the printer actually does what your model specifies, by chance! Yikes!)

When modeling for SW, I create virtually every part with a separate mesh. (Here I must respectfully offer the opposite advice from BillBedford.) In my Scarab and Scarab 2 models, there are more than 20 separate meshes, each carefully overlapped to create a continuous solid object. The downside of this is that I pay a bit more because overlapping material is double-counted by SW. Currently, they compute cost by summing the volumes of all meshes as an approximation to the true total material volume.

The upsides are: first, and most importantly, I can adjust each segment of the beetle independently just as if it were a real creature, which has helped enormously with the model refinement process, allowing me to quickly turn Scarab 1 into Scarab 2. Second, I'm lazy, and Boolean algorithms in Blender still frequently produce non-manifold problems that I hate debugging as much as anyone. No Booleaning at all in these models. (I did print signet rings for me and my girlfriend that used a rather elaborate Boolean difference operation to create the stamp portion. Worked fine, but it took forever to fix the non-manifold problems.)

So, my advice: act as though the printer will do exactly what you tell it to, and don't rely on its errors to produce solid models. Use explicit overlaps if you like, or unify your meshes. Both work. The former is more expensive and requires more thinking but is more flexible if you intend to make major changes to the model; the latter is maximally cost-efficient and foolproof but can somewhat inhibit editing later.

But the text and the bars weren't 0.3mm apart, or even close to that (I believe). I mapped the text onto the surface of the sphere outlined by the twisted bars, so there should only have been some very tiny gaps due to computational error.

On the other hand, I guess I did do some smoothing on the bars, which may have reduced them a bit. Just to be safe, I am now mapping the text onto a slightly smaller sphere.

Thanks for all your advice.

Gillian.