Issues Understanding Flatness

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Issues Understanding Flatness

    I have been trying to figure this out for literal hours. I'm trying to understand how PC DMIS calculates flatness so I can defend it to colleagues and prove it trustworthy.

    My issue is it doesn't make sense. My min and maxs are the EXACT same except for one is positive and one is negative. No way do I trust that. I don't understand where PC DMIS gets the deviation from, it doesn't make any sense. I took a total of 33 random points. I created a plane, leveled and origin-ed in the Z to that plane, then constructed the Flatness. I read somewhere that if I turn on textual analysis on the flatness and then create a location in the Z and turn on textual analysis of the same plane it should be the exact same min and max between the plane and the flatness so long as I level and origin to the plane first. But they're way different and again for the flatness my min and max are the same in the positive and negative which can't be possible. For the plane location I'm getting a max of .002639 and a min of -.003453: the xyz of the max is .666883, 1.250651, .002369 and the xyz of the min is .648687, 1.493985, -,003453. So the deviations for that make sense, it's the z values of the xyzs of the max and the min. For the Flatness, the max is .002711 and the min is -.002711. Why are they so different from the location max and min? In the plane, none of the deviations seem to correlate to the xyz values or anything else. The max and min of the flatness plane are the same as the location so I don't know why I'm getting two separate deviations. I don't understand the deviation the plane is giving me, for the min it's saying a -.002711 deviation but the z value is .003453 and the max is a .002711 deviation but the Z value is .002639. Where does PC DMIS get the deviation of .002711 from? That's what I don't understand and can't figure out and it's driving me crazy. My understanding is that the deviations should be the same for the location and for the plane but they aren't, they aren't even close. If I do the math using the max/min of the location of the plane I get a flatness of .006092 but the Flatness PC DMIS is giving me is .005422. I know the difference isn't huge but I want to know WHY I'm getting a different deviation for the flatness xyz values than the plane xyz values. Where does PC DMIS get a deviation of -.002711 when the actual Z is -.003453? How does that make any sense? Am I missing anything?

    I hope that makes sense, I'm kinda frazzled after trying to figure this out for hours. I feel like I'm so close to understanding how PC DMIS creates the flatness and I'm so close to convincing my coworkers to trust the dimensions given but I don't understand the deviations and I don't know why I'm getting the same number in the negative and the positive for the Flatness call out.

  • #2
    www.c m m guys.com/forum/showthread.php?t=35909&highlight=tangent+plane

    Check out the naughty place (remove the spaces in the URL stupid ********* don't like linking to c m m guys.)
    Last edited by VinniUSMC; 07-29-2020, 01:19 PM.
    "This is my word... and as such is beyond contestation."

    Comment


    • ThePudds
      ThePudds commented
      Editing a comment
      What's wrong with the website? Anything particular? Or just the admins here don't like us talking about the website?

    • Matthew D. Hoedeman
      Matthew D. Hoedeman commented
      Editing a comment
      they don't like the fact that more and better answers are available there.

    • ThePudds
      ThePudds commented
      Editing a comment
      Matthew D. Hoedeman Ahh I gotcha. I just created an account. I'll use both when I can. I like learning.

  • #3
    Originally posted by VinniUSMC View Post
    www.c m m guys.com/forum/showthread.php?t=35909&highlight=tangent+plane

    Check out the naughty place (remove the spaces in the URL stupid ********* don't like linking to c m m guys.)
    That is helpful thanks. My brain is absolute mush at this point and I want to make sure I'm understanding correctly. With the plane location, it's taking ALL those Z points into consideration when leveling but in the Flatness itself it's only taking the 3 worst ones and leveling to that and that's where the difference is? So my deviation between the two should be and always be different and my min and max will be the same number because it's making 2 parallel planes the exact same distance apart as close together as possible while still keeping all the points I took in them?

    Comment


    • gt143744
      gt143744 commented
      Editing a comment
      So Vinni, are you saying that the Flatness creates a tangent plane? I understand how flatness works but some of my colleagues don't trust the flatness measurements from the plane because they're "too good" and to get them to believe me I have to be able to explain why subtracting my max/min of a plane doesn't match what PC DMIS is calculating as a Flatness. They're currently trusting a dial indicator over what PC DMIS is giving me so if I can't explain exactly what PC DMIS is doing and how it gets the plane and such they won't believe me.

    • VinniUSMC
      VinniUSMC commented
      Editing a comment
      No, it does not create a tangent plane. Tangent plane is a construction fitting option. If they're measuring with a dial indicator, they're probably measuring parallelism more than flatness. As mentioned below, the flatness calculation on the CMM is limited by the amount of data it is given. But, unless your colleagues have a flatness setup (indicator sticking up through a plate) instead of the typical sweeping of the surface opposite the surface that sits on the granite, then their numbers aren't right either.

    • gt143744
      gt143744 commented
      Editing a comment
      Thanks Vinni. I know they're not getting flatness and my boss does as well and still they argue with me because the PC DMIS numbers are "too good". I don't think I can get them to trust me on this one.

  • #4
    Flatness is the width of two parallel planes without respect to orientation that fully envelope all points on the surface. The vector of the zone doesn't need to be the same as the vector for the plane, if it is it's purely coincidental. If an orientation constraint were to be applied, it would be profile of a surface.

    Comment


    • VinniUSMC
      VinniUSMC commented
      Editing a comment
      You're right, jerk.

  • #5
    The difference between what any CMM will report and what an indicator report boils down to how many points the CMM takes and the location of those points. Indicator guy is going to sweep the surface. If an area is seen that looks like a hill or a valley that spot will be explored more closely. Indicator is literally taking trillions of points very quickly.
    Typical touch probe CMM program is taking a few dozen discrete points at fixed locations and can quite easily miss both Mount Everest and the deepest ocean trench.
    Typical analog scanner will be dragged across the surface part picking up thousands of points. Better situation, but the scan path is fixed in place.
    Laser scanner will grab millions of points but my understanding is the accuracy isn't all that good.
    CMM's are designed to measure parts with perfect form. They will excel at measuring a ground part but have problems with a rough sand casting. It can't report on what it doesn't touch and the paths that are generated aren't intelligent so that the CMM can zoom in on area's of interest.

    Comment


    • gt143744
      gt143744 commented
      Editing a comment
      Very true. I might have to see if we have scanning capabilities for our machine, however it might be easier to concede defeat on this one and get posts so we can measure flatness manually.

  • #6
    to gt143744,

    I did read you post and I would like to say that your looking at the problem in very complicated way. In your place I would do:
    1) Identify what problem you solving. If I do understand well problem is mismatch of flatness of plane between CMM results and manual equipment.
    2) I would check points which was used to construct plane. In simple words since you have 33 points, I would output T (sheet metal axes) or Z if K=1. In simple words the highest T + the lowest T = your flatness
    3) If the highest T + the lowest T is NOT equal results of manual equipment then:
    a) check surface, if you visually can see defects then move you points where defects are,
    b) if visually you can't see or feel any defects then check manual equipment on very fine surface.
    4) If the highest T + the lowest T is equal results of manual equipment then you did something wrong in pc dmis.

    In other words when you do have complicated problem, best way is to brake down complicated problem to less complicated problems. In your case I would say you have big complicated problem with one plane which is better to brake down to 33 less complicated problems.

    Comment


    • VinniUSMC
      VinniUSMC commented
      Editing a comment
      2) is wrong. Flatness calculates fit without consideration of the orientation of the plane, independent of any alignment. Your T values will very likely never match flatness.

    • gt143744
      gt143744 commented
      Editing a comment
      Thank you. Some of the parts we're measuring are so small and fiddly there's no way to get manual sweeps of them without just crushing them. I think when we do our manual flatness checks we're measuring parallelism and not flatness. It's just tough to try and prove this to coworkers to convince them that we can trust the results we're getting when they're sweeping things with indicators and talking about how their numbers are higher and mine are "too good to be true".

    • Darius11
      Darius11 commented
      Editing a comment
      TO VinniUSMC

      Well, flatness is calculated directly from hits (points) and T is easiest and good indicator of flatness, position and shape plus in T output is easy to do right.
      In case of gt143744 that is only one way to understand and to explain to his coworkers what exactly giving CMM. By having flatness 0.005422 inch (or 0.138mm) is isn't difficult to confirm results by manual equipment. Just remember that to be right isn't enough, need to be capable to prove it.

  • #7
    https://www.pcdmisforum.com/core/image.php?userid=10785&thumb=1&dateline=1491486219
    #6.1
    VinniUSMC commented
    Today, 05:23 AM
    2) is wrong. Flatness calculates fit without consideration of the orientation of the plane, independent of any alignment. Your T values will very likely never match flatness.

    What Vinni said is correct. Do NOT use T Values to consider Flatness. T values are considered from Orientation of your part. Maybe I can simplify this. Imagine you take random points in the middle of space and you start connecting the dots and it creates a plane connecting the dots Still in the middle of space. T-Value is going to relate to your alignment. When checking Flatness nothing else exist. Just like RandomJerk stated now you create a 3D box, no particular origin, still in space, Flatness boundary is a reaction to those given points, 99.99% of the time Your Max and Min will always be equal because your boundary adjust to that particular plane,

    Untitled.png


    Untitled2.png

    That boundary relates to those points only, nothing more nothing less.
    Last edited by KIRBSTER269; 07-30-2020, 12:00 PM.
    (In Memory of my Loving wife, "Ronda" who I lost March 7, 2016. I love you baby.)
    They say "Nobody's Perfect." I must be Nobody.

    Comment


    • KIRBSTER269
      KIRBSTER269 commented
      Editing a comment
      Yeah it's quite often machinist will run indicator over the surface while laying it flat on surface plate, The only accurate way to check it, is to turn it upside down put it on 3 points and run an indicator from underneath. They see you checking it from the top and say "You check it that way" Just tell them "it flips it virtual, shut up and get back to your machine and fix that shi_t"

    • Darius11
      Darius11 commented
      Editing a comment
      From my experience those who are very confident in their true have lack of knowledge and experience. After statement "is wrong" I do see that some persons would like to judge before understanding of meaning of what was said.

    • KIRBSTER269
      KIRBSTER269 commented
      Editing a comment
      gt143744 Word of advise my friend, do not use T value for Flatness, There is a reason why they are 2 different callouts. Flatness doesn't relate to other features, T value is a single form of Profile. Just another person trying to convince others that they are not wrong. In old post when I was wrong, I left my post up and admitted I'm wrong. For people who might make the same mistake or ask questions, hoping all questions to be answered for the next person. That person took my picture, and added values, which is a great example, those points have no values in flatness, again as I stated, I connected the dots to form a plane IN SPACE, Plane has no value.

  • #8
    The CMM is not capturing the same data points as the indicator. The CMM is taking 33 fixed points chosen at random. The indicator is taking trillions of points and is being semi-intelligently guided by the inspector. The CMM will give a general idea of the waviness of the surface at that point density but it is doubtful that it will find the extreme highs and lows of the entire surface. The only shot it has to capture most of the surface in a reasonable amount of time is some sort of high density scanning. Don't remember where or when I heard this but I've been told that the CMM will average picking up only about 70% of the true form error. In my experience that is a pretty accurate statement.

    CMM's are not the right tool for every type of measurement. Form measurement in particular has issues entirely due to point density and the lack of intelligence in the software that will allow it to find the extreme points on a surface.

    If the CMM would show the flatness being out of tolerance, I would believe it, although I wouldn't have the 'true' value. If it shows in tolerance it's a crap shoot.

    Comment


    • KIRBSTER269
      KIRBSTER269 commented
      Editing a comment
      That's why I scan that shi_t. because it takes trillions of points

    • gt143744
      gt143744 commented
      Editing a comment
      I might have to look into scanning. I'm going to see if I can get some posts ordered so that I can go check the flatness by hand. For what we need it for PC DMIS is usually close enough and when I cobbled together a way to actually check flatness it was only off by maybe .001 or .002 which isn't horrible given our tolerances. But I will keep the 70% in mind and I probably won't use PC DMIS for form unless I've got a wide tolerance.

  • #9
    As per given example.
    Flatness can't exceed 0.004 inch beside that I would expect flatness exactly 0.004 because the max T is 0.003 and two points next to it are min T with deviation -0.001. It doesn't meter what orientation of surface, if as in example max point ant two min points is together as in given example, flatness will be min + max. As per given example plane would be twisted. Right side have 0.004 total deviation (left only 0.002), so I would increase amount of points in a right side. Now if as per given example CMM shoving 0.009 flatness, CMM program is wrong if indicator showing 0.009 indicator is wrong. If CMM would show flatness as example 0.003 I may accept it.

    When complicated problem on table, it should be broken down to simple steps.

    222.jpg

    Comment


    • #10
      Another way of looking at this would be to measure the flatness using first principles and to mark the deviations in various places - making sure to include the highest and lowest points. Then set the part up on the CMM and measure a plane, probing the same locations that you marked. Next, create a constructed BFRE plane from your hits, making sure to set the fitting algorithm to MIN_SEP. Level to this plane - you should really rotate to an edge also but for this test just levelling should be OK - now report flatness and turn on textual analysis to see the deviation of each point. The deviations should match what you got manually. Keep in mind that they may not EXACTLY match depending upon the uncertainty of measurement associated with each technique. The uncertainty value for your CMM method is fairly easy to calculate (look at your calibration cert) but for the manual method there are a lot of unknowns - resolution of the DTI being used, flatness of surface plate you're performing the test on etc.

      With respect to the MAX & MIN values - these will ALWAYS be equal and opposite because that is what the MIN_SEP (or tchebycheff) algorithm does. It constructs a plane that equalises the max and min deviations and it this algorithm that both the ISO & the ASME standards say must be used when evaluating form tolerances.

      As many people have previously stated, you can not do a "like for like" comparison between a manual (first principles) measurement and the CMM unless you are taking the exact same points with each method and the measurement uncertainty associated with each method is comparable.

      Comment


      • neil.challinor
        neil.challinor commented
        Editing a comment
        For the first principles measurement, this video shows how to do it. He starts off checking parallelism but then shows how to properly check flatness from about 2 min 45 SEC in. https://www.bing.com/videos/search?q...6FORM%3DVDMHRS
        Last edited by neil.challinor; 07-31-2020, 08:51 AM.

      • gt143744
        gt143744 commented
        Editing a comment
        Thank you! That all makes a lot of sense. We don't really have a great way of measuring flatness manually so I'm stuck between a rock and a hard place. It's hard to tell if you're hitting the highest and lowest points on a CMM but we don't have a precision surface to place a part on for a dial indicator method (and our surface plate hasn't been checked in 40 years). Once we do get one though I'll perform your experiment and see if my numbers correlate.

    • #11
      Here is a great example did a little one for you. I took 4 hits and did TValue for each hit based off my alignment difference between the 4 points shows is .0002 some people would say that's the flatness, but lets be honest this is a different type of parallelism, or single profile because when you make an alignment T values will relate back to your alignment. now if I connect those same dots, and create a plane, and callout flatness, it ELIMINATES alignment, features, location, space, and time only it exist to ITSELF and based off of that (4 points not enough information) flatness shows zero, Anyone who uses T- Value for Flatness is no different than a machinist laying a part on a surface plate and running it across the top saying he's checking flatness, when in turn he's checking parallelism or TValues.

      Capture2.JPGCapture.JPG
      Last edited by KIRBSTER269; 07-31-2020, 12:59 PM.
      (In Memory of my Loving wife, "Ronda" who I lost March 7, 2016. I love you baby.)
      They say "Nobody's Perfect." I must be Nobody.

      Comment


      • VinniUSMC
        VinniUSMC commented
        Editing a comment
        Paul has a new user name?

      • KIRBSTER269
        KIRBSTER269 commented
        Editing a comment
        Rhetorical questions to KIRBSTER269. Are you old guy who knows only one right way which was learned in time of apprentice and any other way is completely wrong. Or you are young guy who learn the way how to do job and now time only doing job and don't have any time to stop and think.


        Darius11 just old enough to know when someone doesn't know what they are talking about and will defend it at all cost to try not to be wrong. I guess VinniUSMC doesn't know what he is talking about either? Are you that Fake News I hear about all the time?

      • Darius11
        Darius11 commented
        Editing a comment
        Wow, KIRBSTER269 you must be very beautiful to yourself.

    Related Topics

    Collapse

    Working...
    X