to be confident or not on a result...

So how would one go about making a validation program, per the specifications described in ISO 10360-2?
I'm specifically interested in ISO 10360-3, regarding CMM validation when a rotary table is the 4th axis.
I've never trusted the rotary.

So how would one go about making a validation program, per the specifications described in ISO 10360-2?

If you can, use a Koba-step twice a year, 19 locations mini, all the steps, 3 reps, and write the result in the same file.
Calculates the average of deviations, the standard deviations and use it as cmm part of uncertainties.
Use a "standard part", which represents what you measure usually, and measure it, and here again, look at the average of dimensions and std dev.

I'm specifically interested in ISO 10360-3, regarding CMM validation when a rotary table is the 4th axis.
I've never trusted the rotary.

You can do the same with a "standard part" !



One more thought, the uncertainty can be divided in three parts, the cmm, the part and the method.
Here, we used to say that the cmm is for 1%, the part for 10% and the method takes the rest ! (algo, number of hits, repartition, reading and understanding the blue print.....).
I would add that the designer is responsibble of 80% of them !

Hmm, an interesting concept to consider.
I will look into this.

So according to your calculations Mr. Calculus what % do you trust your CMM?

I would not be surprised if he came up with the algorithm for trust.
I would actually be more surprised if the number it spit out didn't have a billion decimal places...

KIRBSTER269 , InspectorJester :
a few hikes later (), I would answer that a % is not the right unit for an uncertainty.
Saying the uncertainty is 1% would be 0.01 mm for a length of 1mm, and 10 mm for a length of 1 m... I hope the cmm is better than this !
In my case, i believe that measurements uncertainties are within ±10µm for parts size around (0.5 m)^3, taking into account the cmm, the part and the method, and describing them in the report (number of hits, algorithm....)
To give an example, measurind a Koba step in 19 locations, 52 steps, 3 repetitions give a standard deviation on deviations around 1.2 µm.
If you measure the distance between a point and a least square plane without tacking into account the flatness in the uncertainty, you can be wrong.
An easy way to see the uncertainty is changing the algorithm evaluation on the same COP.
Another solution is repeating the same measurement at the same location, and in differnet locations on the cmm, and each time, looking at the average and the std dev... It can be very surprising !
If you have a large standard ring (around Ø 250 mm), just measure it 10 times horizontaly and 10 times verticaly and look at the average circularity, you should see a little difference, not a lot µm, but the gravity is here, so the uncertainty also .
Sorry for the µm sellers...

I would expect nothing less from you Jeffery

One more thought, the uncertainty can be divided in three parts, the cmm, the part and the method.
Here, we used to say that the cmm is for 1%, the part for 10% and the method takes the rest ! (algo, number of hits, repartition, reading and understanding the blue print.....).

You forgot the operator and the environment (unless you include them in "method").

I would add that the designer is responsible of 80% of them !

That seems to be the metrologist view... I would lean a bit more in the designer's favour, and say 78%...
(as long as we are allowed to take numbers out of the air)
:-)

You forgot the operator and the environment (unless you include them in "method").

I would say that method includes operator, and cmm includes environment, but only in a metrology room, not in a workshop.

That seems to be the metrologist view... I would lean a bit more in the designer's favour, and say 78%...
(as long as we are allowed to take numbers out of the air)
:-)

I grant you 75%, it's 3/4 of the way !
Just a thought about this value, look at the number of metrologists in a GD&T classe, and the number of designers.
Compare those numbers to the total number of metrologists ans designers, you should be convinced that a metrologist has often a better level than a designer !!!!!
Here in schools, designers learn how to draw a 3d part, but they rarely learn how to dimension it.
I'm surprised to see a lot of L±t on a blue print, where a localization would do the job, and simplify the measurement (for older versions).

We actually manage to induce [some] companies to send their designers to our GD&T courses. But they are few, and far between...