Philosophy of CMM Software and CMM

[True Position]

Don't forget that it could be your CAD data not matching the drawing.

Other options: If the hole is quite small, the retract distance / clearance distance to start the scan could be entirely outside the hole, I've had that before. Especially if your default to a 2mm retract distance / 5mm clearance.

They will be in your probing strategy under clearance data.

Try reducing both to .5mm - 1mm and turn the probing speed down a bit. I do this a lot when the probe size is a large percentage of the hole size. For example a 3mm probe in 6mm hole.

Good luck.

 

[Tonycat]

Hi Zancky,

again for the post, I used to program years ago in the old Geopak3, was told back then by the guy who wrote it "Windows is too volatile, and we won't use it!" how times change!

I'm sourcing work back in Quality Engineering again after a gap of a few years (running the ex-wife's business & making her - not me- rich!), and I am soon to go to Mitutoyo for retraining, so your threads are a gift.

Best wishes

Tonycat

 

[mboucher]

-1) writing programs or use "learn mode"?
Personally I have promised to cut one finger to anybody in my lab. each time they try to use "learn mode" (so they have max 10 trials)
Once a techician from Mitutoyo asked me the reason why I was so against "learn mode"
My answer was: "First are You able to elaborate a program in "learn mode" to measure a part with an accuracy about 2 microm wherever is the part, whichever is the operator?.
Second how You can optimize the program (approcing distances, speed, movements) in "learn mode"?
Third how you can compensate the algorithm bugs of Your system in "learn mode"? (I will explain after that topic)
Forth I know trigonometry,some basic other languages and programming tecnique, I can read a drawing, why I can not apply my knowledge?
Which are Your point of view and Your experiences?

-2)How You start your program, i.e. alignement of the axis and origin to the part.
Normally I spend from a third to half of the working time for alignments looking for part simmetry (luckly we are measuring deep drawing parts so there should be a revolution axis)


-3)CMM touch probe, European point of view on measuring (mean value) and Taylor principle (every point inside tolerances).
let's consider just a bore with diameter Ø12±0.1mm for a length of 20 mm. Standard CMM results will give You the mean value and the roundness of the cylinder (mean value is the way used to calibrate the machine). Asking for inner material or outer material diameter will sometime give You strange results (see how roundness changes respect to mean value one). therefore CMM touch probe seems to fit european point of view about measurements and have some problems with Taylor principle. Which are Your point of view and Your experiences?

-4)Touch proble single point bug. Try to measure a inner diameter of a known calibrated ring. if You use the circle or cylinder feature the discrepancy may be 2 microns. Now try to measure the same ring using point (edgepoint). Well if you plot the data on XY diagram You will find that at 45°, 135°, 225° and -45° the points are closer to the center of the circle while they perfecly on the circle every 90°. It seems it comes from a bug into the system. When the touch probe touches the surface it allocates the point coordinates like touching moving along X or Y whatever it is the actual direction vs surface normal. If You change the refence axis according the the direction of approcing to the surface the mistake vanishes. How many other bugs do You know?

-5)I have used CMM to measure wall thinckess on a cylinder like part. Well it took one month to establish the best fixture and the related program but at the end the correlation with the customer laboratory was better then 2 microns. Have you any strange application of CMM?

-6)I have taken the choice to ask to Mitutoyo to perform calibration so my customers are happy (traceability etc..) but I think that 95% of accuracy, while measuring a real part, depends on how I have written the program. the influence on temperature, humidity, etc is a secondary problem. Just to give an idea. Some years ago a customer asked me to measure the minimum and maximum inner diameter point to point on a cylinder like part with nominal diameter of 6.04 mm and wall thickness of 0.15mm. First I said "Sorry but at the moment I'm not able to do it". He said "ok we know a certified/accreditated external laboratory that can perform the task, they are actually calibrating our gauges". After one month we have the results: the certified laboratory declared that the minimum inner diameter was on 5 samples between 5,98 and 6.00 mm. I called them and I asked "how can you explain me that You have found 5.98-6.00 but a Ø6.01 mm plug gauge passes frely through the parts?". "May be there is a mistake somewhere" was the answer.
Do you know something similar?

-7)I would like to know how the machine calculates the features (interpolation algorithms) as asking for microns even the philosophy behing the machine software may have an influnce. What do you think?

Addressing #4 about manually touch probing of points. Are you making sure you are driving on the normal vector to the surface? When probing manually this can introduce cosine error if you are not careful to drive on a proper vector. This why I'm interested if you have tried the same test when using a CAD Model to measure a point. Do you get the same results? There is 3 articles on the CMM Quarterly website about direction vectors written by Richard Clark. They are very informative. You can locate them by clicking on the Past Article tab

COSMOS, MCOSMOS, GEOPak3 are European written softwares, written in Germany so they do follow the same approach that many european softwares follow.

You can write HLL, higher level language, in COSMOS to work around some of your issues or as a self check.

Mark

 

[prototyper]

I would like to start a thread about CMM.
I'm using a small Mitutoyo's one with Renishaw's touch probe and camera system
the handbook with the machine it is not enough for good performance, training from dealers not to be mentioned, help from Mitutoyo most of the time lacking. therefore I learned by myself how to use it, writing scripts, testing (some failures), asking help to some friends and so on. Now I have found this forum and a lot of very good technicians so I suppose it is time to to share my experience and find some suggestions and help
some the topics I would like to discuss with You:

-1) writing programs or use "learn mode"?
Personally I have promised to cut one finger to anybody in my lab. each time they try to use "learn mode" (so they have max 10 trials)
Once a techician from Mitutoyo asked me the reason why I was so against "learn mode"
My answer was: "First are You able to elaborate a program in "learn mode" to measure a part with an accuracy about 2 microm wherever is the part, whichever is the operator?.
Second how You can optimize the program (approcing distances, speed, movements) in "learn mode"?
Third how you can compensate the algorithm bugs of Your system in "learn mode"? (I will explain after that topic)
Forth I know trigonometry,some basic other languages and programming tecnique, I can read a drawing, why I can not apply my knowledge?
Which are Your point of view and Your experiences?

-2)How You start your program, i.e. alignement of the axis and origin to the part.
Normally I spend from a third to half of the working time for alignments looking for part simmetry (luckly we are measuring deep drawing parts so there should be a revolution axis)


-3)CMM touch probe, European point of view on measuring (mean value) and Taylor principle (every point inside tolerances).
let's consider just a bore with diameter Ø12±0.1mm for a length of 20 mm. Standard CMM results will give You the mean value and the roundness of the cylinder (mean value is the way used to calibrate the machine). Asking for inner material or outer material diameter will sometime give You strange results (see how roundness changes respect to mean value one). therefore CMM touch probe seems to fit european point of view about measurements and have some problems with Taylor principle. Which are Your point of view and Your experiences?

-4)Touch proble single point bug. Try to measure a inner diameter of a known calibrated ring. if You use the circle or cylinder feature the discrepancy may be 2 microns. Now try to measure the same ring using point (edgepoint). Well if you plot the data on XY diagram You will find that at 45°, 135°, 225° and -45° the points are closer to the center of the circle while they perfecly on the circle every 90°. It seems it comes from a bug into the system. When the touch probe touches the surface it allocates the point coordinates like touching moving along X or Y whatever it is the actual direction vs surface normal. If You change the refence axis according the the direction of approcing to the surface the mistake vanishes. How many other bugs do You know?

-5)I have used CMM to measure wall thinckess on a cylinder like part. Well it took one month to establish the best fixture and the related program but at the end the correlation with the customer laboratory was better then 2 microns. Have you any strange application of CMM?

-6)I have taken the choice to ask to Mitutoyo to perform calibration so my customers are happy (traceability etc..) but I think that 95% of accuracy, while measuring a real part, depends on how I have written the program. the influence on temperature, humidity, etc is a secondary problem. Just to give an idea. Some years ago a customer asked me to measure the minimum and maximum inner diameter point to point on a cylinder like part with nominal diameter of 6.04 mm and wall thickness of 0.15mm. First I said "Sorry but at the moment I'm not able to do it". He said "ok we know a certified/accreditated external laboratory that can perform the task, they are actually calibrating our gauges". After one month we have the results: the certified laboratory declared that the minimum inner diameter was on 5 samples between 5,98 and 6.00 mm. I called them and I asked "how can you explain me that You have found 5.98-6.00 but a Ø6.01 mm plug gauge passes frely through the parts?". "May be there is a mistake somewhere" was the answer.
Do you know something similar?

-7)I would like to know how the machine calculates the features (interpolation algorithms) as asking for microns even the philosophy behing the machine software may have an influnce. What do you think?


sorry for my english and typing mistakes, but let me know if you are interested on it and your opinion

I've just come across an interesting article which ties up with some of your points.

https://www.qualitymag.com/Articles/Feature_Article/24f7995196c38010VgnVCM100000f932a8c0____


(I have no connection with this publication.)

 

[zancky]

Addressing #4 about manually touch probing of points. Are you making sure you are driving on the normal vector to the surface? When probing manually this can introduce cosine error if you are not careful to drive on a proper vector. This why I'm interested if you have tried the same test when using a CAD Model to measure a point. Do you get the same results? There is 3 articles on the CMM Quarterly website about direction vectors written by Richard Clark. They are very informative. You can locate them by clicking on the Past Article tab

COSMOS, MCOSMOS, GEOPak3 are European written softwares, written in Germany so they do follow the same approach that many european softwares follow.

You can write HLL, higher level language, in COSMOS to work around some of your issues or as a self check.

Mark

using the Mitutoyo QVpack5, I type program codes (well it would be better to say scripts). I'm sure to move along the normal vector as first I take the real data (not drawing data) to fit the circle, them I move the origin reference to the circle center, I repeat circle fitting and origin set up. After in the code it is requested to the probe to move to point i, i.e. Radius=XX, Angle=YYi, with Zeta=ZZ. The contact point is searched along the YYi angle direction, therefore normal to the surface, and print the contact radius and angle cohordinates.
With this procedure You will find the max discrepancies at 45°, 135°, 225°,-45° due to wrong compensation of contact point of probe sphere surface .
If, before moving the touch probe from the origin position (=circle center), you turn everytime the X axsis by YYi angle, then you move to Radiuss=XX Angle=0 and Z=ZZ (exactely at same points like before in the absolute reference system) and the contact point is searched along the new X-axsis(=0° angle direction), now the right value of contact radius will be found whichever is the absolute reference angle.

I hope this will explain what happen.
if anybody would like I can attach the script. Please let me know

 

[zancky]

Hi Zancky,

again for the post, I used to program years ago in the old Geopak3, was told back then by the guy who wrote it "Windows is too volatile, and we won't use it!" how times change!

I'm sourcing work back in Quality Engineering again after a gap of a few years (running the ex-wife's business & making her - not me- rich!), and I am soon to go to Mitutoyo for retraining, so your threads are a gift.

Best wishes

Tonycat

If You need help or some funs please don't hesitate to contact me
see Has anyone ever been harmed by a CMM (coordinate measuring machine)? thread on this forum

 

[zancky]

I have a question for American CMM users or those who are applying ASME Y145M.
I have read a James D. Meadows' book about geometric dimensioning and tolerancing (I have no relationship with him or his website/company) and he has got not a very good attitude vs CMM. He tries to explain that as long as ASME is applying Taylor principle and requires reference plane/cylinder to be found on the highest/lowest points (simulation of counterpart reference plane etc), CMM can not be useful as the inside algorithm is based on least square error (i.e. "mean surface") and can not investigate all the points.
In Europe we are lucky, most of us are using ISO standard where is not requested the Taylor principle (only few times maximum material principle) so CMM can be used quite in every situation of geometric tolerances, asking to the CMM to find the reference planes by itself. ASME point of view seems to be "check if it fits" while the european one seems to be "check how the part look like" and it is more process control oriented.
Now the questions: is it true You have a lot constrains using a CMM living with taylor principle?

 

[True Position]

I can only speak for Zeiss CMMs running Calypso, but here go's.

Calypso offers both the Least Square evaluation as well as the Maximum Inscribed / Minimum Circumscribed / etc type evaluations. If you check a true position of say a hole to another hole at MMC, the CMM will default to checking with the 'functional' evaluation, assuming you are having the CMM default to least squares for diameter/etc. (It actually will ask if you would like it to create an additional item with the alternate evaluation to keep you correct to Y14.5M.)

I've sat in a couple of meeting on this topic with customers and after some explaining generally they want the least squared result, especially if you have a scanning head. If you explain that it's going to filter your results to the three most extreme points out of however many you use and evaluation position/etc from there, they understand why the LSQ evaluation gives much more repeatable useful results. A very tiny scratch/tool mark can wildly adjust the true position with those 'functional' evaluations.

Basically an obnoxious theoretically issue, but mostly theory since in practice you will use least squares for repeatability.

 

[zancky]

I can only speak for Zeiss CMMs running Calypso, but here go's.

Calypso offers both the Least Square evaluation as well as the Maximum Inscribed / Minimum Circumscribed / etc type evaluations. If you check a true position of say a hole to another hole at MMC, the CMM will default to checking with the 'functional' evaluation, assuming you are having the CMM default to least squares for diameter/etc. (It actually will ask if you would like it to create an additional item with the alternate evaluation to keep you correct to Y14.5M.)

I've sat in a couple of meeting on this topic with customers and after some explaining generally they want the least squared result, especially if you have a scanning head. If you explain that it's going to filter your results to the three most extreme points out of however many you use and evaluation position/etc from there, they understand why the LSQ evaluation gives much more repeatable useful results. A very tiny scratch/tool mark can wildly adjust the true position with those 'functional' evaluations.

Basically an obnoxious theoretically issue, but mostly theory since in practice you will use least squares for repeatability.

I agree with You,

1) More or less I suppose all CMM machines have got Maximum Inscribed / Minimum Circumscribed / etc type evaluations but LSQ is the best we can apply especially if we are dealing with Cp Cpk and R%R evaluation. (if we are measuring a diameter we can add the roundness to the control chart in order to avoid any concern about assembly constrain).

2) LSQ method gives more usefull information to the production, i.e. for a circle You can determine if You are out of tolerance due to the mean value (wrong target set up) or due to roundness, i.e. finishing ( set up optimization needed)

3) about reference datum I have in mind that every cylinder in the drawing is tapered in reality so closer to a truncated cone and from experience any defect on the cone may give a very tilted axis if you consider the maximum/minimum inscribed cylinder


sometime I'm wondering if it would be better if every customer use the taylor principle in the drawing (so it will be clear the functional constrains and how to control final product with the functional maximum tolerance range) and the organization design departement translates it to ISO standard for production set-up and control leaving the taylor principle only to some easy requirements.

 

[True Position]

I am with you completely. If you're worried about getting lobed shafts/bores, add a roundness spec, don't cheap out and when you get out of round dimensions start talking about the Taylor principle.

If I am checking a dimension and it seems that I might want to use a 'functional' evaluation, I've found customers would rather have the LSQ and the roundness error with an explanation of what that roundness error can cause in assembly/etc. I had a hole that wouldn't accept a gage pin, when I measured the hole it was oval, so I printed a form plot to show the customer the issue. Much clearer then trying to explain multiple evaluations to half a dozen engineers.