Poor accuracy in CAM

[anthony.a.douglas]

I am trying to make some parts to within about 10 microns.  Pretty reasonable for proper machining.  I thought fusion 360 CAM was suitable for real manufacturing.

And yet when I do my toolpathing and then simulate the result and use comparison view, and use the info screen validation section, where you can hover the mouse cursor over a point and it tells you the absolute distance between the model surface and remaining stock, it shows very poor accuracy.

The cusps are consistently at least 30 microns in depth. Even a the trough of a ball mill's tool path they are 6 microns or more too big. 

If I use the contour toolpath to do a vertical cylindrical surface, the surface ends up 50 microns too big, no matter what I set the pathing tolerance to, if I use smoothing, anything. I set the tolerances to 0.001 and it still makes no difference.

I noticed when I zoom out the numbers in the validation section increase.  But zooming in past a certain point does not help.

I also noticed that increasing the density of parallel finishing paths does not reduce the apparent cusp height.  What?

It's just a total lack of accuracy in many different ways.  Is this software even capable of real accuracy required by real manufacturing?

[anthony.a.douglas]

https://a360.co/2xBGG75

there is the file... sorry if the above is a bit abrasive, I just am a little exasperated.

Here is a test with a simple cylinder drawn natively in fusion (the other was in imported iges and I thought that might be an issue)

It's the same problem, arguably worse.  It's 50 microns off at the highest points!  If this was a bushing or something, it would be totally non functional.
https://a360.co/2XJ82D5

Even if I crank the pass tolerance to 0.0001, and use smoothing, it has no effect whatsoever.

[seth.madore]

Fusion has one area that lacks in accuracy, and that is the Simulation. Posted code is spot on accurate, or at least as accurate as your toolpath setting will allow for. Can you share your file or at least a sample so we can see in greater detail what you are dealing with?

As a point of reference, I'm typically working down to 3 to 5 microns in a lot of what I do in my own shop. At that point, I'm finding my machines to be the weak link (Mori and Kitamura) and not Fusion. Granted, I'm not doing a lot of surfacing, mostly Contour and "simple" work 😂

[anthony.a.douglas]

here, I posted the files above, but so you have the links in an email, here they are. Sorry if the above is a little abrasive, I just got kind of exasperated. 

https://a360.co/2xBGG75

https://a360.co/2XJ82D5

[seth.madore]

In your second example, the circle, the posted code is exactly 1 micron off, depending on what post processor you use. I suspect this is a rounding error out to the 4th or 5th decimal place. Unless you are on a Kern, that's not going to be a measurable issue. A post tweak might be able to tighten that up, but I'm not entirely certain about that

On your first part, it's still unclear WHAT you are trying to machine and what the end result is. As I mentioned in the other thread last week, Fusion surfacing performs best when you have the bodies modeled properly and selected in the Setup. 

You have the offset faces, are you machining a cavity? You have the solid elbow piece, is that also going to be part of what you are machining in either Setup? What's the function of the 10 offset faces?

On your Parralel toolpath, do you intend on actually finishing the model with a flat endmill? This is a poor choice, as you WILL have heavy cusps unless you are running a very small stepover. A bull nose will get you by, or ball nose would be the best choice.

What machine and control are you using?

[anthony.a.douglas]

I'm using LinuxCNC,

The blue object is being made out of compacted sucrose. There is to be a
block of bonded graphite around it.

It works like this, full disclosure:
A layer of graphite is deposited, as a block several millmeters thick. In
that block, a cavity is machined.

Then, sucrose powder is packed into that cavity, filling it. This forms
the downwards facing surfaces.

The sucrose is packed in so hard it bonds together into a solid, chalk like
material.

Then the sucrose is machined. This forms the upwards facing surfaces.

This is repeated for several layers, to build the object.

The blue bent cylinder is just a test object. I am trying get Fusion
working with a test object so I can proceed to real objects.

The process is called shape deposition manufacturing, and it is in it's
infancy, or we would have dedicated CAM software. For now, I have to try
to find a package I can shimmy into the role.

Most of my problems come from not being able to get the software to only
touch the tool to where needed and not go elsewhere.

This is what the offset surfaces are for. I could not find any other way
to *both* make the tool overlap the surfaces, ensuring machining is
complete, but also not "stab" down around the perimeter of the surfaces.
That's the main problem I have. Contact point boundary sounds like it
should be a solution from the description, but it does not work.

The finishing with the flat end mill was just an intermediate step, I will
use a ball nose as you describe later, or instead of the flat end mill. I
was just experimenting.

[anthony.a.douglas]

But like I said, I had these problems many times when doing normal 3d machining, so I think I should be able to find a solution somewhere...

[ant.amanzimtoti]

The fact that Fusion is a pay for product and insists on being online, is already the start of many headaches. There's alot of greed regarding this osftware ie, cost, restriction, service/support and for what it's worth, it is never accurate. No matter the machine PP or sender. Shame 

[seth.madore]

---

@ant.amanzimtoti wrote:

The fact that Fusion is a pay for product and insists on being online, is already the start of many headaches. There's alot of greed regarding this osftware ie, cost, restriction, service/support and for what it's worth, it is never accurate. No matter the machine PP or sender. Shame 

---

There's a whole lot of inaccuracies in this post. I run a job shop professionally (in addition to my Autodesk duties) and I'm knocking out parts with .0001" tolerances. Any accuracy issues you're seeing are likely down to machine and method of manufacture.

And greed? Oh sure, CAD/CAM for $500/yr is greed? True, you don't get 5 axis or true 4th axis motion, but...it's $500.

[brianyUL99V]

We just spent $11,000 renewing our Sigmanest license for the year and Fusion performs way better. Not to mention the $5000 for Solidworks so the parts can get drawn...

[magnus.brandtPB4X8]

We got 4 licens of fusion and run som fancy machines like mikron dmg and kern micro for about 2 years now the precision of the software is pretty god at least down to about 5 micron and if you tweak the post and make the right tolerans adjustmen in the software there is no problen to mill down to about 1 micron if you got the right machine...