Flatness on CMM vs surface plate

[[Th...]]

So I'm checking a part and I get flat to .0045 on surface plate with indicator but I can't get any worse than .003 on Calypso by scanning. I tried different points density, speed, evaluation, different probes.

flat2.PNGflat.PNG

[[Ju...]]

One side of the part may not be as flat as the side the CMM is scanning.

[[Th...]]

Yes, I tried with outlier/filter and it only changed a tenth or two

[[Da...]]

To check flatness on a surface plate, You need to support the plate on three points and check the deviation between those three points on the same side as those three points... Does that make sense ?

[[Th...]]

Yes, I leveled off three points by jack screws. Even had another inspector to double check the flatness.

[[Da...]]

Any chance it is rocking in the CMM? Can you put it on three shims? I see you have an XT, try reducing the gage pressure to 100 mn.

[[Ow...]]

Are you clamping the part the on the CMM in anyway causing distortion?
Has the surface plate been calibrated to be flat within its grade level specification? I've actually had to replace or resurface granite surface plates that are used a lot, they do wear.

[[To...]]

In my opinion, I don't believe leveling a part on 3 jack stand provides a truly accurate flatness check. At least in a correlation with the cmm. The plane on the granite may be level at the 3 jacks but the deviations are not from the derived minimum feature plane created from the max and min points.

[[Th...]]

Thanks for the suggestions. I still can't replicate the flatness on the CMM. The surface plate is calibrated to be flat within the spec. I ran the part with no restraint and lower the measure force to 100 mN...still nothing.

[[No...]]

Please sign in to view this quote.

Exactly what I was about to say.
Flatness has no datum reference, but by placing the part on a plate you create one. The plate controls two degrees of freedom (spatial alignment) and with the indicator you measure perpendicular to that datum. It's more like checking parallelism than flatness.

[[To...]]

Please sign in to view this quote.

The surface plate has no factor in measuring flatness. In order to check flatness using 3 jack stands on the granite plate, you would need to adjust the level of the part to a point that when you ran the indicator over the entire surface, the max and min deviation were equal, i.e. +.0015 and -.0015 which would .003 flatness. This would probably take several attempts to complete successfully.

Screenshot 2022-08-11 054900.jpg

[[Ma...]]

Once you handle firm fixture ( or using some magnets to hold it firmly ) then flatness can be affected with tip size and measuring force.

Certifications should follow some rules - aka path, tip size, point density ... to get repeatable and stable results.

[[Er...]]

Getting 0.0045 on granite and 0.003 on the CMM is a good correlation in my opinion.

I would suggest marking the part up on the granite with zeros and high and low points. Then replicate similar points on the CMM. Use the zeros to create a plane. Measure the other points and see how they compare to the granite.

The good thing about granite is I can "hunt" for the highs and lows. Whereas the CMM has fixed points.

A bad thing about granite is where I zero can affect the flatness reading. An extreme example was a long rectangular part made from coil steel. On the CMM it was clearly shaped like a banana, but barely good. On the granite, with one zero at top middle and the other two at the bottom corners, the part looked like it was bowed up on the middle bottom edge and drooping at the top corners and clearly bad.

[[Ri...]]

I've not had a problem correlating flatness from CMM to Plate. Sometimes, I simply move my three points to another area of the part.

Something I noticed from the Op's post.
The Form value is 0.0029 with a MIn -0.0016 and Max 0.0013

I'm failing to see how Flatness results in a value greater than the Form? (Nearly twice the Form value).

[[To...]]

Please sign in to view this quote.

OP said they got .003 flatness with cmm.

[[Ow...]]

The 3 pad concept is just to ensure the side you're setting the part on, that may not be flat, doesn't have any tip or rock to it while the part or the indicator is being moved about to check flatness of the opposing face. I've checked thousands of parts on a granite surface and the CMM over the years and never had any issues getting the same result on both methods.

How thick is the part, could it be distortion?
What kind of indicator instrument are you checking the flatness with? Could be cosine error with the indicator tip not being aligned with the part correctly, especially if you’re using a digital height stand with a traveling axis perpendicular to the column or if you’re checking it standing up on its side with a traveling height stand with the column not being square to the table, the ole two barrel column Mitutoyo height stands are notorious for becoming out of square.

[[To...]]

Please sign in to view this quote.

Hey Owen, I didn't mean to imply that you couldn't get the same results but there is a possibility that you won't. I've included an exaggerated application where we need to assume that we have leveled the 3 points off but the results won't be the same from granite plate to cmm. In this scenario, if we work at getting the deviations more balanced, we will get closer to the same result.

Screenshot 2022-08-12 124420.jpg

[[Da...]]

Please sign in to view this quote.

Try making a flatness characteristic and removing all filters and outliers, then using CAD->Evaluation to visualize the min and max points, and check that you are seeing the same ones on the plate. It is possible that you are finding a local minimum on the surface plate, but not a global minimum result, and calypso is finding a better median plane.

Use the same Cad->evaluation tool to look at the flatness, and change your fixturing. If supporting the part differently changes the shape (but not necessarily the resulting value) then your workholding is distorting the part. Try the same on the plate. I have seen heavy parts change under their own weight, especially with large areas removed, which seems to be the case here based on your scan path. Put a small weight (~5% part weight) between the jacks and see if a stationary indicator moves.

Have you tried different indicators? You might have a marginal unit.

Are you using an indicator tip the same size as your probe? It is possible (but unlikely) that you are getting some mechanical filtering and hiding small wavelength features on the CMM.

[[Ri...]]

Please sign in to view this quote.

Oops.... 🤣

[[Re...]]

Probably not the case, but I would make sure that the part is de-magnetize so that magnetism isn't effecting either the indicator or cmm.

[[Th...]]

This is a basic cooper plate about 3/4'' thick. I've tried two different height gages and indicator and rotating part to zero off at different area. Today we got another part where Zeiss showed flat is .0036 and I'm getting about .006 on plate. I confirmed with the Mitutoyo CMM at .0037.

[[Ow...]]

I could be wrong, just trying to help but, It sure sounds like you're checking parallelism on the granite, not flatness.
The difference between flatness and parallelism is often misunderstood, I've seen a lot of machinist check parts the wrong way over the years.
If there is any way you could put the part up on it's side (standing up) and then check both faces and then check the parallelism, my guess is that the parallelism equals what you are getting as flatness on the granite. Even if you could only scan say 80% of both faces because the method holding upright might prevent access for the full 100%, it should show close to what you're getting and then look at the flatness plot of both sides and then the parallelism plot and on screen.

[[Er...]]

Please sign in to view this quote.

Have you tried marking the Zeros, highs, and lows to compare with the same points on the Zeiss?

[[Br...]]

When you measure the flatness using a CMM, the output is a computerized algorithm where it optimizes all the points as close as possible to the theoretical perfectly flat plane, then give you the distance from the two furthest points apart. With calypso you pick the algorithm. Least squared will take the distance of every point squared, add them together, then iterate until the sum of the squared distances is the smallest value. the Minimum zone algorithm (i.e. "chebychov") minimizes the maximum point distance.

As you can see, computing a scan of hundreds or thousands of points into an algorithm to optimize your flatness output would be virtually impossible to replicate on a granite surface plate. As many have already suggested, the industry standard for surface plate inspection is using a 3-point method. This is not going to be as optimal as a computer algorithm. What you should find is the CMM outputting a slightly smaller value than the surface plate inspection. If you are ever seeing a larger value with the CMM, it is probably an indicator there are some issues with outliers, or the CMM is covering areas that were missed with the surface plate inspection.

[[Br...]]

Please sign in to view this quote.

Owen is exactly right, If you are just laying the part directly on the surface plate and measuring the top surface with an indicator, you are measuring Parallelism. You need to remove the influence of the bottom surface that is interfacing with the surface plate from the equation to measure flatness.