Auto Model Orientation for best print

[blackt1ger]

So, with all this Pokémon Go craze, I decided to print out a Pokeball. And I came back to an interesting situation: when printing half a sphere, which is the best orientation?

a) open side down - most amount of support needed, top part of sphere suffers from layer granularity, that is, small curves relative to the bed over a large model creates a 'step'
b) open side up - less support needed (significantly less then a) still has layer granularity
c) open side horizontal - ????? Probably the best, but takes the longest amount of time.

Although this is just a general case, it would be nice to handle much more complex model. I envision this like a "driving planner" --- Three choices available - fastest, shortest, best gas mileage, for example. Simply3D would take existing print parameters and give you multiple options. I bet some solution would be not-so-obvious, like 75 degrees from the horizontal.

You could use the 'auto-support' parameters to rotate the model and calculate optimum rotation angles for least filament, 'speed' options to determine fastest print rotation and 'layer' options to determine best quality rotation.

The reason this would be great is because it would help inexperienced people to get good results and be able to compute various options faster than experienced people.

(ps. if you like this idea, send me a free t-shirt, men's medium/large)

--Ken

[dkightley]

This sounds initially as it might be a good idea.......but in my opinion impossible to code into the software.

Take how to determine the least amount of filament based on rotational position. With a rotational resolution of 10 degrees, the slicer would have to do 36 x 36 x 36 slicing exercises to determine the one position that would use least filament..that's 46,656 goes at slicing. For a one degree resolution make that over 46 million slicing runs! The same would go for determining the optimum position for speed....and how on earth can the software determine the quality of the print???

I'm afraid I believe we'll all need to rely on the current way of doing things....practise and experience!

Can you tell I'm not in to this Pokémon thing??

[blackt1ger]

This sounds initially as it might be a good idea.......but in my opinion impossible to code into the software.

Take how to determine the least amount of filament based on rotational position. With a rotational resolution of 10 degrees, the slicer would have to do 36 x 36 x 36 slicing exercises to determine the one position that would use least filament..that's 46,656 goes at slicing. For a one degree resolution make that over 46 million slicing runs! The same would go for determining the optimum position for speed....and how on earth can the software determine the quality of the print???

I'm afraid I believe we'll all need to rely on the current way of doing things....practise and experience!

Can you tell I'm not in to this Pokémon thing??

While I do agree with some of what you said, finding the best solution does not have to be that difficult. Going back to my analogy about driving, obviously there is a near-infinite number of routes that can be taken from point a to point b. It is easy to start to see when the results start getting worse and abandon further consideration. It would not be required to find the absolute best, but you can heuristically get several localized optimums.

Hey -- I'm trying to challenge these guys. Simple things like "how much time is left" on a print seem to be ignored, so maybe we need to raise the bar a bit.

[dkightley]

Going back to my analogy about driving, obviously there is a near-infinite number of routes that can be taken from point a to point b. It is easy to start to see when the results start getting worse and abandon further consideration.

But we're not talking about comparing going from a to b along different routes when you know where both a and b are.....we're talking about comparing going from a to a via an unknown number of points with going from b to b via a totally different number of points before you've established where a and b are and how many points you will need to pass through and what their locations are! You can only do the comparison once you've resolving both routes fully....or nearly fully if you constantly compare then whilst resolving the second route.

I'm not saying it's impossible to develop what you're suggesting.....but surely its counterproductive introducing something that takes an hour to work out how you can save five minutes!

[Lagbert]

This article http://hpcg.purdue.edu/bbenes/papers/Vanek14SGP.pdf discusses a method of minimizing required support structure.

They rotate the model to minimize the number the number of over hangs. I can't recall if it's in this article or another one I read, but just rotating the part to minimize its projected area can go a long way to reducing over hangs.

Following the rotation they do a tree style support structure.

Printing support material is such a time killer, being able to have simplify auto position the model to minimize supports would be great.

[Paul_W]

This is a feature that Meshmixer has - under Analysis - Orientation. That is one of the things I use Meshmixer for, and since it's free - why not?
It will orient it for the least support material, then just export the STL in that position, or use the angles in S3D.

[Caall99]

+1 to auto orientation capability (to minimize supports) being built into S3D.

Also +1 to tree-like supports! Much more efficient..