Looking for blend feedback

[BjoernJohnsson]

I'm currently working on blend, and I would love to get feedback from everybody who has tested it.

As I understand it, it is currently inconsistent, and basically failing on too many examples. It would be really great if you could post any failing examples you have in this thread, or send them to me directly, so we can make sure that we fix the issues.

It would also be good if you could post examples of where it gives a bad result, and an explanation of what you would want instead in those cases, or cases where you think it gives a good result, that we should try to get in more places.

Thanks,

Björn Johnsson

Sr. Software Engineer for HSM CAM

[Laurens-3DTechDraw]

@BjoernJohnsson

Awesome stuff you are doing on this!

It is kind of the mother of all finishing toolpaths so tricky to get right. But so useful in so many places.

The new bidrectional step-over check box is really usefull. Awesome work there.

It solves one of the main issue's that users spoke about when it was first released for testing.

I've got a couple examples.

Blend Test 1.sldprt

Has too many retractions and has the issue of the guide curves having to be on the surface to be machined. Due to that the faces will not look as good, and produce code that is harder for the machine to execute at speed. Since there will be constant moving of the Z-axis because of that.

See this picture and the attached part.

I have made a blend test part and some operations don't produce what I would expect/want it to do. And some don't generate.

I attached the part too.

And I send in a part I can't share here to the development team directly on 12-09-2017 that shows blend creating a weird path.

Keep up the good work!

It takes longer than we users want. But that's nothing new, we want it yesterday and flawless, but we do know that it doesn't work like that.

[cj.abraham]

For those who have not used blend, here's a quick example of how blend is supposed to work.

Rail selections:

Drive surface selections:

Right now the selection is restricted to faces that can be driven by two rails as shown above. In the future, it may be possible to drive blend from a single rail which would aid in allowing blend to generate toolpaths on "non-quadrilateral" geometry.

[LibertyMachine]

What product is this being developed for currently and which will have the greatest chance of success? I'm not having much luck in Fusion..

[al.whatmough]

@LibertyMachine We identified a UI issue in Fusion that is preventing the current BETA of Blend from working.  We are in the process of correcting this.

If this was not the case, you should be able to test in all products.

[LibertyMachine]

Forgive me if I'm being obtuse... What level of product? Should I download the Inventor HSM 2018 beta release, or should it be working in 2017?

I did download the 2018 and give it a whirl, but encountered two issues:

1) my license ## isn't recognized, so I'm in a 30 day trial

B) the Blend failed on me anyway

[al.whatmough]

@LibertyMachine  For Fusion it will be in the base product as 3X and when we release the 5X version it will be in Ultimate.

For Inventor HSM we only have ultimate now.

As for the case that failed can you attached the file and share a screenshot?

[LibertyMachine]

Sorry, let me clarify:

Is this only expected to be working in Inventor HSM 2018 (Beta deployment) and HSM for SW (Beta Deployment)?

[scottmoyse]

I thought I would be mean to it, and use some T-Splines. Sure enough it failed. See attached.... also the curve selection is a bit cumbersome in this example... it doesn't pick a favorable guide curve at first selection.

It also fails in HSMWorks with the same error... but the selections are SO MUCH EASIER! 

[scottmoyse]

What is Use bidirectional stepover meant to do? I see Laurens' example above, but I don't really understand it, other than seeing the difference in the number of points... I also don't really understand the significance of the old style vs flow style comparison.

[scottmoyse]

It looks like this retract is unnecessary:

I've attached the file... but also note that Blend 2 generated here... but the same toolpath in Inventor HSM didn't. Both files attached.

[scottmoyse]

Why is Multi-Axis Blend it's own toolpath in HSMWorks.... but a checkbox in the passes tab of Blend in Inventor HSM?

[Rob.Lockwood]

attached file from earlier today that I spent a few minutes attaching some blend toolpaths to. The first job is what I wound up running on the actual part, though it was a quick project to get a probe mounted, so the operations are far from ideal.

Each operation has notes..

I'm finding Blend quite reliable for groups of clean quadrilateral based surfaces, even if they have occlusion from the main surface, which is a big plus.

But multi-axis version is failing unless the tool can be tilted clear to the surface normal along the path..

And rails still REALLY need more flexibility, as otherwise non-quad surfaces are pretty much useless.

[Laurens-3DTechDraw]

@scottmoyse about the bi directional, make a part that looks similar and you'll understand the difference. It's about what edges it divided up to make extra passes.

[scottmoyse]

---

@Laurens-3DTechDraw wrote:
@scottmoyse about the bi directional, make a part that looks similar and you'll understand the difference. It's about what edges it divided up to make extra passes.

---

Turning that setting on vs off fails for me... see attached.

[BjoernJohnsson]

About the bidirectional step-over, the difference is in how it guarantees the step-over.

Looking at two curves on the surface, with bidirectional on, it guarantees that there is no point on the first curve that is farther than the step-over from the second curve AND there is no point on the second curve that is farther away from the first curve. 

With bidirectional off, the guarantee is that either there is no point on the first curve that is farther away from the second curve OR there is no point on the second curve that is farther away from the first curve.

 It mostly makes a difference where these curves (blue) are at an angle to the boundary (black) of the surface. If you look at the picture, with bidirectional on, it guarantees that the distances marked by the red and green lines are both less than the step-over. With it of, it guarantees that one of them are. For the right part of the example, it doesn't matter, as the curves are perpendicular to the boundary, and thus the distances are almost the same. For the left part, where the curves are at a steep angle to the boundary, it does make a difference, because the red distance is much longer, and this will result in the curves being pressed together. 

[scottmoyse]

---

@BjoernJohnsson wrote:

About the bidirectional step-over, the difference is in how it guarantees the step-over.

Looking at two curves on the surface, with bidirectional on, it guarantees that there is no point on the first curve that is farther than the step-over from the second curve AND there is no point on the second curve that is farther away from the first curve. 

With bidirectional off, the guarantee is that either there is no point on the first curve that is farther away from the second curve OR there is no point on the second curve that is farther away from the first curve.

 It mostly makes a difference where these curves (blue) are at an angle to the boundary (black) of the surface. If you look at the picture, with bidirectional on, it guarantees that the distances marked by the red and green lines are both less than the step-over. With it of, it guarantees that one of them are. For the right part of the example, it doesn't matter, as the curves are perpendicular to the boundary, and thus the distances are almost the same. For the left part, where the curves are at a steep angle to the boundary, it does make a difference, because the red distance is much longer, and this will result in the curves being pressed together. 

---

@BjoernJohnsson do you have a dataset or example you could share with a couple of toolpaths showing this please? I'm still struggling to wrap my head around which curves and geometry your sketch and description refer to. 

[Laurens-3DTechDraw]

@scottmoyse

It's about the edge it encounters.

With bi-directional it makes sure it looks if the steps between the end points of the passes don't lay further apart than the given stepover. No matter if you look at it vertically or horizontally.

With that off. So what now is the "Normal" setting. It just divides the edge up between the guide curves while making sure it never goes beyond the set step-over.

Attached is a part where you can see the paths become very different in some cases.

[scottmoyse]

Perfect @Laurens-3DTechDraw that makes perfect sense to me now in combination with @BjoernJohnsson's description. So it will be interesting to understand why it is failing in the dataset I created and attached here that's similar to yours.

[scottmoyse]

@BjoernJohnsson if we can only select two drive curves, can it be limited to just two selections in the dialog, by having two single entity inputs boxes instead of one which allows as many as you can select?

 I'm guessing there will maybe be support for more inputs here in the future version of this toolpath?