I feel dumb, need help with flatness.

[[Mi...]]

"GD&T Flatness is very straight forward. It is a common symbol that references how flat a surface is regardless of any other datums or features."

This has always been my understanding of flatness, very straight forward, and only checked to itself, low to high points.

So can someone help me understand why the alignment of the part has an affect on flatness?

I wanted to just take a quick look at a surface, so I made a plane and a circle to locate the part. Rotation in space = plane, Z origin = plane, X/Y origin a simple circle in the bore, the part in rotationally symmetric so no Planar rotation. My flatness result is .0017"

So I changed the alignment, added an additional circle in the bore, created a 3d line using the 2 circles. Rotation in space = 3d line, Z origin = plane, X/Y origin = 3d line, no planar. Result is .0007", which is much closer to what I can see manually running and indicator over the surface on a granite table.

If flatness is only checked to itself, why does my alignment affect the result? Flatness is so simple that I feel really dumb that I can't wrap my head around where the extra .001" is coming from, because clearly I am introducing it.

 

Thanks.

[[Ze...]]

Are you measuring the exact same area of the plane each time with the same measurement strategy? Are you taking enough points to adequately define the plane? 

BTW... if you are putting the part on a granite plate and running an indicator across it, you're checking parallelism, not flatness. 

[[Mi...]]

Please sign in to view this quote.

It's the same plane used for the alignment. I only have 1 plane and 2 circles in the very quick and basic program.

Please sign in to view this quote.

Yeah, I do know that, I just figure if the flatness is at .0017", I should see some movement on the indicator even checking on a granite.

I could nearly duplicate and see the .0007" results checking this way however.

 

If I use the plane1 & circle1 alignment, I get .0017". If I use The 3d line1 and plane1, I get .0007". Same plane is being measured and used for the alignment.

[[Ma...]]

Was that plane scanned twice? Or how you tested your theory about alignments modifying flatness results?

Is it replicable - aka you run 1st as plane, then 2nd as 3d-line, 3rd as plane and still those results?

[[Mi...]]

Please sign in to view this quote.

I'm using a different part today, so the results will vary.

Using the bad alignment of Plane1 & Circle1 as pictured before, I take 4 points to grab the plane, then 4 points in the bore for Circle1. Start CNC, then a 1250 point Plane1 is scanned, then a 500 point circle (semi-circle technically to avoid a keyway) is taken. Flatness result is .00198"

I change the alignment to 3d line1 & Plane1, take my 4 point Plane1, then a 4 point Circle1, then a 4 point Circle2 over 2.5" deeper into the bore. Start CNC, 1250 point Plane1 is scanned, then the 500 point Circle1, then another 500 point Circle2. Flatness result is .00066".

Since you asked for it, I now switched BACK to the bad alignment of Plane1 & Circle1, 4 points, 4 points, CNC, result is .00199"

At no point during this has the part been moved, but if I chose to 'clear existing results' and keep current alignment, it will rescan the Plane1 and only Plane1, and the results do not change since it does not retake the new alignment.

 

I hope I explained that clearly enough.

[[Mi...]]

Thinking about this some more, I think I may know what's going on, and I guess it does make sense, hopefully someone can confirm what I am thinking is true.

The Flatness measurement still needs a perfect form or median value to evaluate from, and when I give the program a 3d line for the rotation in space, it places the plane1 at 90° to the 3d line. But when I use the Plane1 as my rotation is space, it is probably comparing that plane to the perfect form or median plane that would be created by the machine itself. So, my CMM stage has some tilt to it, and that is what I am picking up.

 

Would that be correct?

[[Ch...]]

I can't replicate this no matter what I do on my offline seat.  This tells me its unlikely its the alignment per say (more on this in a minute).  I don't think your theory about the machine tilt is what's causing it either.  

But here's my theory:   Since changing your spatial datum in the base alignment is clearly causing a repeatable difference, it's possible you are seeing some sort of probe radius correction error.

 

 

Is this a rough part, like a casting or plastic?  What kind of difference do you see between using filters vs no filters?  I'm just pontificating here, but maybe Plane 1 is too small to establish a reliable spatial datum; maybe that's why you get better results when you use the larger 3d line in the base alignment. 

 

Anyone else got any ideas?

[[Ze...]]

If you're manually probing your features are you cleaning up the nominals/vectors of all the features? The software should be creating the planar tolerance zone relative to the feature alignment & feature nominals related to that alignment. Flatness should be best fitting the tolerance zone. Are you measuring the plane as a separate feature not included in your Base Alignment?

Edited 20 hours ago

[[Mi...]]

Please sign in to view this quote.

Ground surface and ground bore (see picture)

Please sign in to view this quote.

about 1 tenth.

Please sign in to view this quote.

It is small, no denying that. It is taken with a circle path on a 2.000" Ø

Please sign in to view this quote.

I don't understand what you are saying, sorry. I included a screenshot of the feature plane1 window. ELI5, Feel free to RESPECTFULLY talk down to me, if that makes sense.

Please sign in to view this quote.

No, as I mentioned before, I have 1 plane and either 1 circle or 2 circles to create a 3d line., check the screenshots I previously posted to see for yourself. This is a very simple basic program that I threw together to simply check a ground surface flatness. This isn't even a print requirement, just me trying to help the shop troubleshoot some issues.

[[Ja...]]

Can you make a secondary alignment that is either the 3d line or plane (whichever one is not the current base alignment) and set your feature to it and see if the form changes? It shouldn't. If it doesn't I would see how perpendicular your 3d line is to plane1. I suspect you scanned a different area, maybe you got closer to the keyway, and that's why your number changed. I would also look at the actual points and see where your min and max points are moving to based on the alignment, maybe that will clue you in to what is happening. 

[[Mi...]]

Please sign in to view this quote.

I think I did what you asked. The form changes. The bore is NOT very perpendicular to the face, this is the very issue I have been trying to help them solve, there is absolutely face runout on this face when compared to the bore, which is translating into the gear when it gets ground, because they are colleting on a small section of the bore, and resting on the ground face. I know the bore and face not being very perp to each other is an issue, but the parts are already done, I am simply trying to help them understand what has happened.

As far as scanning a different area, its the same 2.000" Ø circle path, centered on the alignment/bore.

And the 2 circle paths for the circles in the bore, are both 500 points, start angle of 0°, then I have it set to start at 120° and scan for 300°. Basically I only scan CCW from 11:00 to 1:00, and I simply place the part of the stage with the keyway aligned at 12:00. I have to use a 3.0 mm long probe to take the 3d line, otherwise I'll risk shafting out. And the keyway is very thin and shallow, so I simply am ignoring it and programming around it, rather than using it for a Planar rotation and I would normally do to give a fullly defined alignment.

[[Do...]]

If evaluating flatness is the primary goal, I would stick with using the plane as the spatial axis and a single circle path as the origin. The 3d line looks to be introducing error in the measurement by skewing the axis ever so slightly, thus placing your planar scan path in a slightly different position than the original alignment. 

To test this, would you be able to add a few more scan paths on your plane at different diameters? Maybe one smaller and one larger than your original path? It would be helpful to see the form of the plane over as much as the surface as you can. 

Ensure the proper filters/outliers and eval method (min. feature) as usual and adjust your speeds accordingly.

From there, if still curious, measure the larger plane again against the 3d line alignment and check the flatness plots. How do they compare? Are the highs and lows in relatively the same locations? Are they the same magnitude? 

[[Da...]]

The alignment does not change the flatness. But you should consider, that if you use filtering and outlier removal, that there might be a number of points for your plane that borders on "filtering works" or "filtering doesn't work". Sometimes the filtering breaks down if there aren't enough points for the filter to work. And since the filtering parameters can change from element to element and characteristic to characteristic, it might be worth to check every element and characteristic for the right parameters, and also check the default protocol for information about the filtering (it tells you when the filter doesn't work).