hexagon logo
Register Sign in
  • Replies 12 replies
  • Subscribers 221 subscribers
  • Views 182 views
  • Users 0 members are here
Options
Related

Cylindrical vs. spherical tip probes on sheetmetal thru holes?

Alex B-shift
I have a disagreement with a customer as to how I'm measuring their part on my CMM (this is the same part I was posting about back last October). I have about 3 dozen 10mm PUNCHED thru holes on a sheetmetal weldment (.050 sheet copper, tin-plated). The assembly is actually 2 layers of sheet copper, and the holes are one layer or the other, with embossments around the holes bringing them to roughly the same datum plane.

In my past, places that I've worked where sheetmetal was measured on the CMM, we always used to use stainless-steel cylindrical "barrel probes" with a radiused end, and left the spherical end probes for measuring machined parts. This was (as I was told) because of the nature of sheetmetal in general giving you a very small 'window of flat surface' to hit accurately with a ball probe, and also because of the unpredictability of 'blowout' from punched holes, or the raggedness of lasered holes. Even a machined hole on sheetmetal only gives you a target area the size of the metal width.

My customer, OTOH, has decided to measure the part with THEIR CMM, using a (I believe) 1.5mm spherical tip probe, and they are wondering why they are getting different results than us. Their CMM guy's take is that spherical probe tips are more accurate, and that somehow using 'pre-hit' before measuring each hole makes z-axis mishits non-existent.

Also, about a dozen holes are filled with M8 PEMS, which he is also measuring on the threads with the same tip (whereas I decided to measure the PEMS around the outer diameters). These PEMS are only being held to a tolerance of TP 040, so it's not like they are 'tightly toleranced'. The thru holes are toleranced to TP's of .010-.020.

So, oracles of the CMM world, who is right? Who is wrong? What would you do in a similar situation? We need to make a single program that we can both use, but we seem to be going at this from different styles and don't agree on much. FWIW, their programmer is a guy with 20 years experience on CMM's in machining, whereas I only have about 14 years (on many different softwares), but I'm also a licensed mechanical engineer and have been involved with sheetmetal for about half of my career.

There were also other disagreements in our programming style; he believes in using only his take on a 'progressive alignment', which in his world means: Measure plane, open alignment box and level plane, then set as Z origin. Close alignment box. Measure X-axis line, open alignment box, rotate line to X+, set as Y-origin, and close alignment box. Measure Y-axis line, etc.... He claims that my 'full alignment' which I accomplish in just one alignment box session (in the order of Level, rotate, origin z, origin x, origin y) the same thing AFTER measuring all the elements is invalid.
  • If your customer is getting repeatable results as proven by a gage r&r then I would go with their method. It's not always about being "right". It's about having happy customers. Make sure your concerns are documented both their and at your facility.

    If your customer is using sample hits around holes and edge points around the outside then I think the method is valid as long as the hits are on the cut surface. Barrel probes are dependent on how square the probe is to the material and how wavy" the surface is.

    How much difference are you seeing between your programs?
  • They are 'right' in using the ball and the prehits. Using a cylinder is ONLY going to be correct IF the shank of the cylinder is perfectly perpendicular to the surface (hole axis) being measured. While you can 'calibrate' the shank of the probe, there is no place in Pcdmis where you can tell Pcdmis that you are using the shank and not the tip of the probe, which is where ALL data comes from.

    As for the alignment, it will 'depend' on what the alignment is done to. If you are measuring a holding fixture that sits flat on the table, then yes, you are correct, the full alignment (manual) followed by a DCC alignment (ALWAYS do a DCC alignment after a manual alignment) is perfectly fine. If, however, the plane you are measuring is NOT flat to a machine axis, then you will get probe comp error if you don't level first before measuring a hole or a line.

    For the 'stepped' alignment he does, it does sound INCORRECT to me. Yes, measure the plane, level & origin to it. Then measure a line, but then you SHOULD level & rotate in the next alignment. Then measure the last feature, then LEVEL, ROTATE, and set origins in the last alignment. It has been proven that Pcdmis CAN SOMETIMES muck up the THEO values in a program if you do NOT have 1 alignment that locks all 6 DOF at once.

    Remember, NEVER inspect a part with just a manual alignment, always follow a manual alignment with a DCC alignment. Always.
  • Here is a graphical representation of the cylinder probe issue.

    BLUE is the part surface and the shear of the pierce
    GREEN is the break of the pierce
    RED is a cylinder probe at the center of the pierced hole, perpendicular to the hole
    PURPLE is a cylinder probe making contact on either side of the hole when it is NOT perpendicular to the hole
    the PURPLE cross-hairs are where the CENTER would be calculated based on the touches of the non-perp cylinder probe touches.

    As you can see, a cylinder probe will NOT give the correct location if it isn't perfectly perpendicular to the hole.

    10 degrees out of perp gives 0.7+ mm of off-location.

  • in reply to kingsld1
    "How much difference are you seeing between your programs?"
    Maybe .05mm on thru hole locations, .10mm on PEM locations.
  • in reply to Matthew D. Hoedeman
    "It has been proven that Pcdmis CAN SOMETIMES muck up the THEO values in a program if you do NOT have 1 alignment that locks all 6 DOF at once.

    Remember, NEVER inspect a part with just a manual alignment, always follow a manual alignment with a DCC alignment. Always."

    Yes, that is what our differences boil down to; He thinks doing piecemeal alignments is somehow better than addressing all 6 degrees of freedom in one single alignment. And, no, I never run with only a manual alignment on any DCC machine.
  • in reply to Matthew D. Hoedeman
    Hmm, that's more than I expected, although I can't see my assembly being more than about 1° out of square. My issue would be more that, in the graphic example you are showing, and say I'm taking 8 hits, 2 would be dead-center, 2 would be on the upper side where the diameter was smaller, 2 on the bottom where the diameter was larger, and then 2 would hit past the corners, giving an even larger diameter, and possibly an incorrect center. With the barrel, I'd get a closer approximation to what a pin check would also give for the diameter, and a center based pretty much strictly on that same pin diameter. I would NEVER try measuring a part which was visibly non-square like in your example.
  • Just wanted to add: I've never really used pre-hit. I know that it will allow me to take a single or multiple hits on the surface around the hole, but how much will it 'adjust' the hits I take INSIDE the hole? Will it just adjust my Z axis to that specific hole? If I have 3 pre-hits around each hole, will it also square my k-vector specific to the hole as well?
  • in reply to Alex B-shift
    3 sample hits ("required" for a hole) makes a plane that Pcdmis will then measure all hits in. If the surface is 10 degrees off, the hits in the hole will be on that same 10 degree angle. Think more like a temporary level alignment that Pcdmis also 'origins' to.
  • in reply to kingsld1
    yeah, those PEM things do NOT have any kind of a concentricity spec on them, the outside isn't a good thing to measure to prove where the inside is at.
  • I did some testing using a cylinder probe and measuring the calibration sphere at different levels. Zero on the sphere then measure a circle at -1 -2 -3mm The results are attached.
    Attached Files
© 2024 Hexagon AB and/or its subsidiaries. Privacy Policy | Cloud Services Agreement