Diameter of a hole, AME, UAME, theory vs reality and other nonsense 1

[sendithard]

Been digging in the weeds of the Y14.5 standard a bit. The more I read, the more I learn, but it feels like I get dumber sometimes.

My most profound question is what is the diameter of a thru hole and a blind hole? Is it the largest pin that will slide thru and largest pin that will go to depth?
---I believe my above statement is correct and that is the UAME(unrelated actual mating envelope). When I was reading that nonsense it just doesn't make sense. The actual mating envelope is the pin, and the unrelated envelope is the theoretical. Seems backwards to me.

Then Y14.5 hits us with 'The actual local size of an individual feature at each cross section shall be within the specified tolerance of size." (pg 18 2009 standard)
---So in theory, you could have a pin that fits within the tolerance of size, but due to form issues(pitting, tool marks) the part could fail if you sectioned the part and measured it. Not to mention you have to measure it perpendicular if the hole had a curve to it show in Fig 2-6 2009.

 

[sendithard]

Bur,

Here are two videos to explain things in more detail. Let me know if you have any questions. More than happy to explain or do another video as this forum and you have been so valuable in my learning of GDT.

https://youtu.be/gFzS7zLZOec

https://youtu.be/17_-J3T75-c

 

[Burunduk]

sendithard,
You have done a great job with the videos. Very helpful in understanding this inspection software's terminology and the process of working with it.

I do think I now understand the process better thanks to your efforts, but let me get this straight:

1. For datum simulation, the method set in the "fitting elements" definition dictates how the scan will mate to the simulated datums - it's high points/least squares for planes or maximum inscribed/least squares for holes...
But the "fitting elements" method is only retained if "Geometry" is later selected at the datum system definition as the "computation mode", and it can be overridden if you select "GDT" which forces a datum simulation according to ISO. Is my understanding correct?

2. In the second video, when you click to add the "main alignment" and select "by coordinate systems" you mention it is just for visualization. You also mentioned a nonmandatory visualization step when you showed the text listing the steps in the first video - which I believe addresses the same thing. So when does the software actually moves the part scan from the state of "pre-alignment" and aligns it with the datums according to the selected method of datum simulation? Does it happen automatically for a geometric tolerance measurement that is related to the ABC "datum system" according to the "computation mode" defined for ABC? Or is it possible to have a measurement related to ABC performed while the part is still in pre-alignment? I've seen inspection reports for geometric tolerances generated by the software that had "pre-alignment" stated beneath the table. Does that indicate a problem?

I appreciate your help very much.

 

[Burunduk]

sendithard,
I'm watching your videos again to let everything sink in.
I have another couple of questions:

3. In the first video, after you created datum A as a nominal auto-feature (blue) related to the nominal CAD, you apply the fitting element, and you get the green plane related to the scan. You say that the datum itself from which measurements are made is the fitting element plane (green plane) and not the (blue) auto-feature plane.
Then you show that there is a difference between them visible.
Are these two supposed to become coincident with each other once the alignment is performed with the condition that datum A is the primary? My understanding is that they should, and we are only able to see them separate of each other when the parts are shown in pre-alignment or in some other type of alignment that doesn't consider datum A or doesn't use it as primary datum. Am I correct on this?

4. Also if the fitting element planes are the datums, what would happen to the fitting element if A was not the primary datum in some datum system you could create? When you create a fitting element for an auto-feature, it has the maximum amount of contact with the scanned surface according to the chosen method (high points or least squares). Then if you use that datum as secondary, it should accept a constraint of staying perpendicular (or at any other basic relationship) with the primary datum, so it should then make less contact with the surface of the scan. Does the program recalculate these fitting elements that are used as datums according to the datum precedence order used in a particular datum system?

Thank you!

 

[sendithard]

Let me begin by saying, Auto Feature is what you create from the CAD and you do a fitting element for the scan. If I don't have a CAD model and just want to do some dirty checks and whatnot, I just click on a scanned surface and go straight into Fitting element. The software is smart enough to know what you are looking for and you can choose cylindrical pickers, plane pickers,etc, and you can even do pencil like area stuff. Seems like you got a lock on this, I just wanted to reiterate it. The green stuff is the real deal, blue is all theory. And the cad is there mostly just to help the program 'find the real surfaces'.

1) Yep correct, since the scanned part has surfaces that become the real datums your choice of least squares as lets say a bottom Datum A surface this part will sit lower than a part if you forgo that selection and use GDT for computation. So if you do a height measurement from this datum it will read .9995" with lst sq and like 1.0005" for GDT.

2) You can create datum structures till you are blue in the face. The software does nothing with them, until you do a GDT like measurement such as Profile or position and call out a particular datum structure. And even then it calculates the position based on the ABC but visually you still see the part and CAD in the pre-alignment state.

From my understanding and stress testing, I don't see the need to ever do what I call a visual alignment to a datum structure unless you want to show a machinist how the CAD and scan are fitting together under an ABC datum structure. Even then I don't need to do this...I can highlight my 'profile of a surface measurement' and toggle it to a heat map and it shows the heat map or a go-no go color map that a machinist would prefer better and it does this using the datum structure even when I'm not visual aligned to said datums.

If you have some tight geometry that you need to make sure what you click on the CAD gets applied to the correct area of the scan, you can draw around these areas and do a quick 'local alignment' Then, b/c you married two areas up very well, the CAD surface click in the Auto Features will be applied more accurately when you do your fitting element. Where I use visual alignment is this....I'm not a fan of the heat maps and go no-go color maps as much as I like the below pic for showing a machinst the profile of a surface callout......I'll just include this in the video instead of a pic.

3) 100% correct. It's cool watching you work thru how it should operate and then look at it and see it does this....I will do a video on this later for my records and for some colleagues. I will show you how in the video the primary CAD blends perfectly into the Scanned primary, but you just had to ask about the other datums didn't you?

4) Okay, this is out of my league here, but I am going to experiment with it and see what I can learn. If I have to I'll build out some high points on CAD and stress it to see what is happened when I know what should happen before the fact. Don't expect and answer soon I will say this, once you do a simple square part with datum ABC the CAD top datum and real datum mesh perfect. Then the datum plane X an Y are perfect to the world and look like lines viewed from the front, but if you unhide the real green plane surface it is angled a little. So I'll investigate how it is navigating this matrix. From what Ive seen everything is well done. Only someone well versed in GDT could ask that question. I assume just like to slide a part on a surface plate up against an angle block for datum B it does the same here, it just will allow you to let this sliding action happen tangent plane style or least squared which would mean the angle block and some of the real part would intersect to some degree.

Good stuff, I'll yell at you later this week....

 

[sendithard]

btw...if I remember in some post you are the engineer or do cad work, if so you need to make up a perfect part and save it as your master step file then alter the geometry in a bad way and save it as _bad_step. Then inside gom load the bad step file and simply go into Operations > CAD > CAD to mesh then export that stl file in the proper units. Then load the perfect step file into a new project and import the bad mesh on top of it...do a prealignment as it is mandatory even though they prob are already prealigned for the most part. Then play around and enjoy the brain damage.

 

[Burunduk]

sendithard,
Thanks again for clarifying all this so well!

If you have some tight geometry that you need to make sure what you click on the CAD gets applied to the correct area of the scan, you can draw around these areas and do a quick 'local alignment' Then, b/c you married two areas up very well, the CAD surface click in the Auto Features will be applied more accurately when you do your fitting element.

This bit sounds important...
So I figure from what you say here that this what any kind of "alignment" done manually may really be important for, in the context of geometric tolerance measurement results ("manually" as opposed to what the "Geometry"/"GDT" datum system computation does). It seems to allow a more accurate recognition of the actual scan surfaces when their determination is done by derivation from the CAD to the scan, and then perhaps the "fitting element" datums are related to the scan more accurately. Doing it with just prealignment is not the same. I guess this may aid in obtaining more reliable and repeatable measurements in certain cases.
An interesting experiment would be toggling between prealignment and local alignment or other types of alignment once the measurements are already made and letting the software recalculate the results, watching whether the measured values in the report change how much, if they do.

I'm not a fan of the heat maps and go no-go color maps as much as I like the below pic for showing a machinst the profile of a surface callout......I'll just include this in the video instead of a pic.

Very interesting, and very important.
I'm definitely looking forward to it.

 

[Burunduk]

sendithard,
I've been thinking a bit more about everything we've been discussing and about what I've learned from you regarding the software, and I think I am coming to a conclusion on what the "fitting elements" are, essentially. I think a fitting element is just a basis for all sorts of computations related in one way or another to a specific feature from the scanned (actual) part. For example, for a flatness evaluation, you create a fitting element of plane type representing the planar surface, and then the flatness tolerance applied to it measures the portion of the cloud of points which is associated with that planar fitting element (probably the area of points that was selected during the fitting element creation). It does not measure the fitting element itself because the element is a theoretical plane and it wouldn't make sense to measure its form as you would get a zero flatness value. Then if you associate the fitting element with a datum system a datum plane is constructed in some way near the fitting element correspondingly with: 1. The associated portion of the scan and its actual form (such as finding the high points) 2. The computation method that was defined for the datum system ("GDT" or "Geometry"), and 3. The precedence of this datum in datum system (primary/secondary/tertiary).
So considering all these factors the fitting element may end up being the datum itself (when "Geometry" is set and the datum is the primary), or very close to it (close as in the difference between ISO and Chebyshev outside for example), or it can be some plane near it.
Thoughts?

 

[sendithard]

Yea, I agree with all you said. For the flatness, you almost really don't need a plane except for the fact you just gotta give it an area to analyze. You wouldn't want to manually box the entire top surface b/c the edges might have rounded areas so you want to stay off them by backing off the edges. Then it just measures flatness deviation against itself and whether the plane is gaussian/chebyshex outside doesn't mean a thing here. One used as a datum plane it would matter just like you said.

The fitting elements like you said are the real deal. You learn this fast when not using CAD at all and only use fitting element. The CAD and auto features make the software parametric and repeatable, I can keep dropping in 100's of scanned parts into one program and boom I have a report. Without a pre-alignment and these cad Auto-features I'd have to drag each individual scan into the program and built out new fitting elements for each part, would take forever.

I have tested position, surface profile, and some other GDT checks with only the pre-alignment, and then grabbing the fitting element with a local alignment. The results were not material(.0002" change) I haven't really focused on this aspect yet. So far the software does a great job grabbing the feature from the real scans. This is also where the person engineering the scan comes in. How accurate is the scan/is the machine calibrated? I calibrate the CT scaner every two weeks and I must be there for 3 hours and then it finishes up an hour later by itself. When I find a situation that needs a local alignment for proper fitting element capture I'll let you know.