Adaptive machining improvement?

[LibertyMachine]

This idea is prompted by THIS THREAD and my own personal experiences.

When Adaptive clearing a large area of material where there is nothing left at the Z plane (almost as if you had faced it) the last few cuts can get quite hairy, especially if you have enough depth of cut or it's a troublesome material. Currently, pretty much the only way to prevent part or tool destruction would be to Adaptive clear to an "island" and then have a facing operation remove that last little bit.

I'd like to think there could be a way to better automate it. I'm thinking a small dialog much like there is for 2D Contour and "Tabs". Select "Island Safety" or some such name and define shape and location. Or, it could be parameter driven, such that it would be 60% of tool width or similar.

Thoughts, comments, criticisms?

[Laurens-3DTechDraw]

I'm not really sure what you mean.

Could you add a screenshot maybe?

[LibertyMachine]

So the email update problem isn't unique to the Fusion forum I see......

It's that sliver that gets left when you adaptive clear an entire face:

[Laurens-3DTechDraw]

Some time ago there was an update to the Adaptive kernel to fix it leaving material in the middle.

So the last cut is positioned so that it actually cuts away all the material. But doing a very deep AP with very shallow Ae you will get lot's a of tool deflection which could potentially cause issue's you can't really blame to the adaptive kernel but more to physics 😛

When a 2D Adaptive is possible a sketch would be a quick way to do this of course but when the shapes get more complicated I understand that isn't as easy.

[LibertyMachine]

Right, I remember that kernel update and things were greatly improved then.

Yes, by going with a deep AP and shallow Ae, there are issues when it gets down to a thin sliver, especially when it loops around and heads back the other direction. It's at that point where I'm often "chuncking" out that sliver and it breaks off, hopefully with the tool unscathed. With the $$ value of the parts I run (as well as tools, cuz carbide ain't cheap) I often will 2D Adaptive clear to an imaginary island (sketch), roughly 60% of cutter width. After the adaptive, I'll do a face operation to remove that section down to where the adaptive left off. This prevents part or cutter destruction, although at the expense of more manual interaction.

What I'm advocating for is an additional dialog, much like 2D Contour has with tabs. If there was one for adaptive facing that left a basic shape of an island, it could possibly eliminate the need for 2 separate operations as well as the required sketch editing

[kb9ydn]

It's been my observation that cutting the sliver works ok until it has to go around the end to the other side.  That's when it gets scary, because the sliver starts deflecting into the cutter instead of away from it.  What the adaptive strategy needs is a way to detect when this condition could occur and then only machine from one side.  Or it needs to disengage from the material before crossing to the other side (this would be better for long slivers)

I wonder if this could be detected by looking for outside corners that are significantly less that 90 degrees?  That's sort of what's happening here.

C|

[Steinwerks]

I agree with @kb9ydn, in that once the sliver or post is too thin the tool needs to skip cutting around it as if it is solid anymore. I've broke small end mills on these sorts of thin "vibrating" features. I think machining from both sides is possible, but not engaging on the ends (second suggestion).

[LibertyMachine]

Thanks for chiming in @Steinwerks and @kb9ydn. Yes, it's that loop around that causes the most issues, although I'd be willing to bet it starts right before it starts to loop around. If they changed the kernel to avoid this, they might have to go with a more aggressive arc-off motion to greatly reduce the chance of the sliver getting sucked in

[Laurens-3DTechDraw]

@kb9ydn @Steinwerks

I would be all for something that detects the thickness and only cuts from one side.

It's something I have suggested in the past and it sounded like a hard thing to implement.

We can see if @martin.dunschen is able to give his view on that.

[al.whatmough]

Solving this in an automatic was would be excellent - we agree.

Essentially what you want is to remove the last 5 passes and leave the upstand and then attack it in a different way.

It seems easy - the experts tell me it is not.  In the mean time, you could model an Upstand.

We have an ongoing internal conversation about this.   Hopefully, the conversations moves far enough along that Rene or Marin can weigh in on this thread.

[LibertyMachine]

Awesome!

Just curious, what does the team have in mind, a tweak to the kernel so it handles the slivers differently, or added options for handling that last sliver?

[al.whatmough]

@LibertyMachine  Both.  For starters, we are working on changing what stock is left.

http://autode.sk/2v9brdG

We have also looked into possibly attacking that final plug in Z slices.

Thoughts?

[LibertyMachine]

Can't view the video, but I like the idea. A "both ways" facing operation down to Z depth would be great, along with some math to define the size of the sliver (perhaps no more than 60% of the cutter width)

[al.whatmough]

WE actually left behind a plug instead of a sliver:

[kb9ydn]

Cool!  This has been needed for a long time.

Video is marked private.

C|

[martin.dunschen]

Hello @kb9ydn, sorry, my mistake, the video (caveat: this is just a preview to try out and discuss this as a possible solution) should now be publicly available.

I took the idea for it more or less from @LibertyMachine, of leaving a central area (island) to be cleared away last. 

I am also still not quite clear about what causes this problem:

- The sliver is not evident when you simulate the stock, everything is removed.

- Tool deflection is obviously always a factor in how much stock is actually removed, but for the last thin blade of stock it won't be the tool that deflects but the stock.

So, why is the blade not being cut off?

Examples where users reported this problem seem to have in common that they use flat bottom cutters with no corner radius. But usage will presumably very quickly lead to a small corner radius from tool wear? Now that would lead to the base of that remaining blade to be stronger and not to shear off so easily, right? But that would also mean that your tool engagement value (Ae) would be smaller that this radius from wear, and that is not the case (typically Ae might be 25% of the tool diameter). 

Please share your thoughts with us here, your input is invaluable to us.

[LibertyMachine]

@martin.dunschen

The problem is rarely that there is material left over after the adaptive toolpath.

Rather, the problem is that the last sliver can be destructive to tools. The depth of cut versus the size of the sliver tends to lead to the sliver getting sucked into the endmill and something is going to break. It's either a $50 tool, or a potential scrapped part.

This isn't quite a problem when you are only taking a shallow adaptive pass, or working with a soft material like aluminum. Bring in a hefty (20mm+) depth of cut, add in some stainless or inconel (shudder) and you are likely going to end up with a broken tool.

So, what some of us are after is mostly a means to not destroy stuff or place excessive wear on tools. Tool cost can be a real expense for a shop

[Laurens-3DTechDraw]

@martin.dunschen

@LibertyMachine

Martin,

You fixed one problem, that is that the silver was being left last year.

What we are talking about is the issue that I made this picture for last time we spoke about.

(The green line doesn't exist in the toolpath now.)

The current problem is when the stock still left is getting thin. Like about below 4 mm and you are cutting 0.5 wide when it then goes round the thin end you’ll hear and see very loudly that it eats the stock. It sort of smashes the tool straight into the very thin wall. From the thin end. Which it shouldn’t. When the material left gets thinner than a certain amount is should start on the thick end only.

@al.whatmough maybe we can make one location to have this discussion?

[martin.dunschen]

@Laurens-3DTechDraw, yes I remember, but thank you for explaining this problem again. 

There is a recent video here from NYC CNC where John still sees the thin blade (he then breaks off using some pliers) - that looks like the thin slivers can still be a problem. In the video it starts machining again at the thick end and does not turn round the top. 

To me it now seems that we might be talking about two separate problems:

- the appearance of thin blades not being removed (as in the video and also here)

- collision of the tool with a thin strip of stock 'head-on' (as @Laurens-3DTechDraw describes above)

[LibertyMachine]

@Laurens-3DTechDraw agreed, this conversation needs to be taking place in one location, not split across two

I've done a fair amount of 2D and 3D adaptive, and i can't say that I've actually encountered slivers being left (in recent months). My money goes to an incorrect tool definition or selection error, rather than looking at the software as the culprit