Wrong bridge detection result

[SoftFever]

Fixed a bug that top surface was mistaken as bridge.
To produce this bug, just set thick top shell layer when slicing Orca cube

Fixes #4648

@igiannakas
I reverted this part of the change. I tested the case you mentioned, it seems fine too
#3285 (comment)

[igiannakas]

Hm that bug is surprising. Have not seen this happen in the past few months with this change active. I’ll test again with the model file from the original issue and also validate that the bridging expansion is still ok.

thank you 🙏🏻

[igiannakas]

Ps. I’ve also noticed in the latest prusa alpha that there are changes to how bridge infill expansion is handled. Haven’t noticed that bug either with orca but maybe something to keep in mind for the future: prusa3d/PrusaSlicer@b133579

prusa3d/PrusaSlicer#9999
prusa3d/PrusaSlicer#10493

[SoftFever]

Ps. I’ve also noticed in the latest prusa alpha that there are changes to how bridge infill expansion is handled. Haven’t noticed that bug either with orca but maybe something to keep in mind for the future: prusa3d/PrusaSlicer@b133579

prusa3d/PrusaSlicer#9999 prusa3d/PrusaSlicer#10493

yeah, Orca has optimized for bridge.
But bridging to a thin parts is still a challenge.
Haven't got the chance to check what has PS changed yet.

[igiannakas]

And I dont think it fixes the issue with this model I'm afraid:

which I would have thought so as the core logic in discover_horizontal_shells is skipped due to the continue statement when ensure vertical shell thickness is set to all, so the re-ordering of the this->process_external_surfaces(); shouldn't affect that code at all...

[igiannakas]

While it does introduce the bridging issue when ensure vertical shell thickness is critical only or moderate

OrcaCube-2.3mf.zip

[igiannakas]

This PR build with Stock Voron profile with 5 top shells:

Test 6.3mf.zip

So that is definitely not the fix I think... Something is causing the top layer to be considered an internal bridge.

A workaround for this is to set the top layers to 6 or 4 resulting in the bridge layer not coinciding with the top surface layer of the (Z) but this is clearly not the solution

[igiannakas]

Replicated in 1.9.1

Replicated in 1.9

[igiannakas]

Same issue with Prusa Slicer 2.7.1:

And latest Prusa Alpha (2.7.3):

So the fixes from Prusa on bridge detection in 2.7.3 won't fix this.

[igiannakas]

Doesnt happen with BBS:

So seems like a bridge expansion defect with both Orca and Prusa when the top most layer is exactly at the same height as an internal bridge supporting a secondary top most layer.

I amended the STEP file in fusion increasing the depth of the Z cut out by exactly 0.2mm

However when increasing the top surface thickness by 1 more line:

Which supports that theory - when the layers are perfectly aligned in that specific scenario the bridge expands fully.

[igiannakas]

More to add:
For some reason when ensure vertical shell thickness is disabled, the top surface is detected correctly. Maybe because the Top surface expands more?

[SoftFever]

I've tested with more cases.
Now I'm getting inconsistent results, lmao.
I agree it's not fixed, and the cause is very unlikely the change that I thought was.

[SoftFever]

bridge seems fine tho?

[SoftFever]

anyway, I will revert this PR first

[SoftFever]

So seems like a bridge expansion defect with both Orca and Prusa when the top most layer is exactly at the same height as an internal bridge supporting a secondary top most layer.

I agree

[igiannakas]

bridge seems fine tho?

The regression happens when the ensure vertical shell thickness is not all ;) I've attached a project file above for your reference on this :)

[igiannakas]

I've been trying to find the source for this with very limited success. The PrintObject::bridge_over_infill() code seems to be working ok, as in it detects the unsupported areas properly (orange lines below). Blue is the expanded bridge region in that code which still is correct, i.e. doesnt fill the whole layer.

But something then expands the bridge and replaces the top surface.

This can be "resolved" by adjusting the expansion top value below but I think that is a bit "hacky" and not the right way.

[SoftFever]

I've been trying to find the source for this with very limited success. The PrintObject::bridge_over_infill() code seems to be working ok, as in it detects the unsupported areas properly (orange lines below). Blue is the expanded bridge region in that code which still is correct, i.e. doesnt fill the whole layer.

But something then expands the bridge and replaces the top surface.

This can be "resolved" by adjusting the expansion top value below but I think that is a bit "hacky" and not the right way. !

I have a hunch that it's probably just because the original algo failed to consider the cases that internal bridge and top surface can be on same layer.
I will give it a try later, will keep you posted.

[igiannakas]

I've been trying to find the source for this with very limited success. The PrintObject::bridge_over_infill() code seems to be working ok, as in it detects the unsupported areas properly (orange lines below). Blue is the expanded bridge region in that code which still is correct, i.e. doesnt fill the whole layer.

But something then expands the bridge and replaces the top surface.
This can be "resolved" by adjusting the expansion top value below but I think that is a bit "hacky" and not the right way. !

I have a hunch that it's probably just because the original algo failed to consider the cases that internal bridge and top surface can be on same layer. I will give it a try later, will keep you posted.

Yeah same feeling here but for the life of me I couldn’t find where to adjust it :) thank you!!