hexagon logo

Feeling Like An Idiot

So I'm running a 7.10.7 with an indexable head using touch trigger probes, checking some really big parts. Occasionally we switch over to check some smaller parts with lots of deep internal features. These are literally scaled down versions of the larger parts. To do so I have to add a 100mm extension between the probe body and CMM head. Recalibrate all the probes in my rack. Update the loadprobe command for like 10 programs using 4X40_Probe to LONG_4X40_Probe as an example. Takes about an hour to an hour and half I guess. I do this when ever its needed which is about twice a month sometimes.

So I had the idea of keeping the extension on the CMM at all times. I wrote a program offline and I was able to clear my large parts by indexing to A45B0 to move in and around my part, safely index down to A0B0 and continue inspecting. It worked offline and I tested it on my CMM last weekend and it worked amazing. I thought I solved my problems of having to come in and remove/install the extension from happening but I was wrong. That weekend when I tested my program and part it failed the some angle dimensions. I thought nothing of it and figured we would just write an NCR and just keep going. Well its been failing those same angle dimensions on every large part we've made since then.

So its been a week and I was doing some troubleshooting today. Cleaning scales, cleaning and re calibrating probes and cleaning parts. Someone suggested to recheck and failed part, then remove the extension and re test a failed part again. Well I did that and it passed the those angle dimensions. I guess the culprit was the extension? I don't understand why it would fail those angle measurements on the large part with the extension but pass them on the smaller still with the extension.

Anyway I had to email engineering and production and quality that the issue was the CMM the whole time and the CNC machine.

Parents
  • I had a customer that had to use long extensions to access tight bearing bore IDs in a deep bore from both ends of the part.

    They had variation in their results and had placed a ring gage on their fixture to measure the this in each index position (A90B0/A90B180) and they would print this result on the report for verification or a measure of how good the part maybe. It was confusing for their machinists.

    I set up a test for them to measure the ring gage, index measure the exact same hit points from the other side, then loop and repeat. It showed that the head was not repeatable at these long extensions.

    I then showed them that they could perform an autotip calibration after they indexed within the program.

    We repeated the above test but included the autocalibration. The repeatability became excellent and a fraction of their tight tolerance, where as without the autocal, it was similar to the tolerance they were trying to measure.

    If you have room on the table to place the calibration sphere so that you can autocalibrate after tip changes with long extensions, this will get around your problem.

  • Quick question, David. When doing the auto calibration on the tips with long extensions in your example, the error in calibration that is removed isn't from the probe/tip build itself, but from the head of the machine? How long of tips should we start trying this with?

    Our machine is a Hexagon Global S. We use the HH-AS8-T2.5 Head + HA-TM-31 + TP 20, and there is very few programs with long tips. But I have seen programs with 100mm and 130mm long probe builds here that don't include a hard extension or something like an EM1 or EM2 due to needing as skinny and long of as a probe as possible. Should we be trying your method in these cases? I haven't noticed any parts being out when these programs are ran, but I also do not always know when they are ran due to us having 3 shifts of regular inspectors.

Reply
  • Quick question, David. When doing the auto calibration on the tips with long extensions in your example, the error in calibration that is removed isn't from the probe/tip build itself, but from the head of the machine? How long of tips should we start trying this with?

    Our machine is a Hexagon Global S. We use the HH-AS8-T2.5 Head + HA-TM-31 + TP 20, and there is very few programs with long tips. But I have seen programs with 100mm and 130mm long probe builds here that don't include a hard extension or something like an EM1 or EM2 due to needing as skinny and long of as a probe as possible. Should we be trying your method in these cases? I haven't noticed any parts being out when these programs are ran, but I also do not always know when they are ran due to us having 3 shifts of regular inspectors.

Children
  • it will depend on extension length and what tolerances you are trying to measure too.


    You can do this repeatability test to understand the effect of the head indexing repeatability for a given extension length.

    At typical probe build lengths, indexing heads are generally sub micron repeatable (otherwise your would not be able to use them). This repeatability is magnified by the extension length as when you perform a single calibration the head is in a certain position but next time it indexes, it will be in an ever so slightly different position contributing to an error. Indexing heads are convenient but not as accurate as a fixed head (all the leitz high end machines are fixed head for example) 

    my customer had a renishaw PH10 head and 200mm/300mm extensions on it for this one part. The auto calibrate increased cycle time but greatly improved accuracy and also removed the need for a verification measurement of a ring gage in both index positions to try and judge if the measurement was good. We also restructured their program so that we only indexed a minimum number of times.