Part alignment workflow on a Haas VMC

[SADIM_cnc1]

Hi @Richard.stubley 

first of all: many, many thanks for your video content, these are pure gems for programmers wannabe such as me.

I subscribed yesterday to  the extensions in order to apply the part alignment functionality to a very simple case I have to deal with.

background:

I have to add milling features to existing PMMA lasercut letters. I used to eyeball the position stick them with double sided tape on toolboard. I would then capture the WCS at an angle with the build in angled WCS probing cycle, and use the deviation angle result in a G68 program rotation. This isn't fast but at least it is reasonably easy for all the prismatic shapes I have to deal with (eg letter N, L, E, etc).

intent:

Now for odd shape such as S, O, Q, etc, I used to stick the letters along a jig to get the proper orientation and WCS positioning. This is now where I'd like to use the part alignment feature. I think this will be easy for such a simple case like this one and a good way to learn the workflow.

Prep:

on the haas I set the setting 261 to FILE so the DPRNT data comes out an actual file

I set the setting 262 to a dedicated folder in the netshare folder I'm using at the shop, eg "my inspection folder"

in fusion:

-I create a part alignment sequence and choose 3 axis with freedom only in X and Y (this what the stock is supposed to do when I eyeball its alignment and location when I stick it on the table)

-I create what I think to be the minimum of 3 points to capture an alignment and see where the part has shifted and how it has rotated. (I plan to place more points for odd shapes.)

- I set the upper and lower tolerance to a very large value as I know the part could be a couple mm off (may be I'm wrong in the way I understand this one?)

- I then post with the haas milling inspect surface post and run the inspection in commissionning mode.

Result at this point:

I have a DPRNT file, but it is empty (0 kb). This is very odd as I did generate an DPRNT file with some data in it in the previous attempts. I just can't reproduce it.

Beside, I had a look at the second or third inspection run in the DPRNT file. Is it ok or not to have it appending with multiple START and STOP or should I aim to a file with only one START and one END?

Would be recommend to change something in the above?

I'm so pumped up to try again first thing in the morning:-)

cheers,

Rémi

[SADIM_cnc1]

Hi there,

I figured out my DPRNT output issue; I left the sequence numbering on.

Now I turned it off I have a proper DPRNT file.

So far so good, I now reinject these data and reprocess to align the part.

I programmed the first op as visual alignment check operation. It is basically a spot drill that hovers above each stock corner to confirm the alignment correction before proceeding further.

Unfortunately I can see that it is way off from the part.

It looks like my workflow is really close to what it is supposed to be but I can't get it to work properly yet.  This is so interesting though, @Richard.stubley   do you think you could put my on the right track for this?

many thanks

Remi

[Richard.stubley]

Hi @SADIM_cnc1,

Taking a look now 🙂

To answer an earlier question.
No, it does not matter if you have multiple starts and ends in the results file. We scan through the results file on import and only take from the last START onwards. This means you can keep appending to the file and it wont effect the import.
 

[SADIM_cnc1 · Accepted solution]

 Hi everyone,

huge thanks to  @Richard.stubley for the brilliant support, including video. This is second to none.

This is the debrief as I find it important to break it down and share with others.

On the Haas control side:

-setting 261 has to be set to "file"

-setting 262 has to be set preferably to a local machine folder such as "user data". This works fine and his reliable. Choosing a USB drive or remote network folder works also, but may not be as reliable for data collection.

On Fusion360 side:

-Richard workflow is perfect and can be followed exactly as shown.

-A part probed will carry as many control points as you choose to use, but they must be projected on the solid to be taken in account with the alignment process. Un projected points can't be computed.

-A 3 axis alignment expected to correct a part that would shift in X+Y and rotate around Z has to be set in "4 axis, Z rotation" to allow the alignment to be calculated.

-with machines not equipped with rotary, the post properties in group1 must set top "no" in order to confirm there are not physical rotary axises available to help moving the part and get the alignment.

-a  setting in Parameter/manufacture/projection search distance may be to small if your part to be probed is way too out of whack. Increasing this value allows Fusion to project the probed points onto the model when they are far out. Keeping this value as low as possible is however a good practice for complex surface parts.

Conclusion:

This is a new workflow to me but is has proven itself to be extremely useful and accurate in order to aquire parts location and orientation quickly and precisely. @Richard.stubley  please edit/add/correct if some of the above is incorrect, and many thanks again for your help, this is brilliant!

kind regards,

Remi

[Richard.stubley]

Hi @SADIM_cnc1, 

Excellent write up, I'm glad we have got you up and running and can see the true power of Part alignment. 

So to elaborate a little more, the main issues we faced were to do with "Projection".

The problem we face is when we do Part alignment we know the part is in the wrong location, so how do we trust the values that are given back?
The answer is projection, I wont pretend to know how it works, that's for someone much cleverer than me to explain. But in short, it looks at where the point was taken then projects that point onto the model and works out where it was actually taken. 
Again because we know the part is wrong, we know where we asked it to take the point on the part is not actually where it took the point.

So to be able to project these points we need 2 bits of information.
Projection bodies: We need to know what CAD geometry you were touching, was it the part model, was it a stock model, was it something you have modelled up separately. We do our best guess but sometime you need to override our best guess. You can do this when you Import results (See pic and the selection "bodies for projection")

Projection Search Distance: This limits the distance that we can project the point for. "why don't we just make this value huge?" I hear you ask. Well you can project onto the wrong surface, so you should be setting this to how far wrong you think the part might be. See Images for where we set this in your preferences > Manufacture > Projection Search Distance.

Looking at the first example here, you can see that there are 2 points "Un-projected" this means that the points could not be projected so in effect do not exists. So the alignment will be calculated without them. 

Here now we can see that I have increased the projection search distance  and all of our points are projected, and have aligned perfectly 🙂

 

Hopefully this is all becoming clear. I do have a series of part alignment videos planned and I will be making intentional errors like these un-projected points.