hexagon logo
Register Sign in
  • Replies 17 replies
  • Subscribers 221 subscribers
  • Views 345 views
  • Users 0 members are here
Options
Related

Scan results different from keyed-in points

TheBradish
I have a part with a very tight profile tolerance (form+location) of .0005. I wrote a program with a model and aligned to all 3 datums.

I keyed in gage points and read the vectors from the model. Part is out of tolerance (which I expected) by about .001, the profile being off center.

Next I added a linear TTP scan along the same section where I probed the gage points. The scan shows deviations up to .018!Confused Same program, same part, same alignment.

What could I be doing wrong with my scan to get a result that differs so much from my keyed-in points?
  • "keyed in gage points"

    These points MAY not lie exactly on the cad data. Don't just find the vectors, FIND the entire point. Pcdmis will take keyed in values and find the closest CAD XYZ nominals.

    If the PRINT is master (I assume this is where the gage points come from), then don't use the cad for nominals. If the cad is master and the gage points on the print are 'target' points, only use cad nominals.

    Also, I seriously doubt that they gave you a list of gage points as close together as the scan points will be, so that comparison is apples to donkey carts.
  • I have a part with a very tight profile tolerance (form+location) of .0005. I wrote a program with a model and aligned to all 3 datums.

    I keyed in gage points and read the vectors from the model. Part is out of tolerance (which I expected) by about .001, the profile being off center.

    Next I added a linear TTP scan along the same section where I probed the gage points. The scan shows deviations up to .018!Confused Same program, same part, same alignment.

    What could I be doing wrong with my scan to get a result that differs so much from my keyed-in points?

    I am assuming you have a manual align and DCC align? maybe the part was moved.

    could it be that you have probe comp OFF? .018 is .5mm
  • "keyed in gage points"

    ...If the PRINT is master (I assume this is where the gage points come from), then don't use the cad for nominals. If the cad is master and the gage points on the print are 'target' points, only use cad nominals.

    Also, I seriously doubt that they gave you a list of gage points as close together as the scan points will be, so that comparison is apples to donkey carts.

    Print is master, but angles are not defined, so I have to use the model for that. After zooming in close, it looks my keyed points are not actually on the model, as you suggest, but scanned points are. Not sure where I went wrong here. After starting a brand new program I have gotten my scans to agree with keyed in points (or at least much closer).

    I am assuming you have a manual align and DCC align? maybe the part was moved.

    could it be that you have probe comp OFF? .018 is .5mm

    I did to manual and DCC. Part was not touched during this process. I kept re-measuring and got the same divergent results.
    Probe Comp is ON. (If only it could be so simple...)
  • Latest quirk: As posted above I started a new program and my scan got a lot closer. OOT by tenths, not several thousandths.

    I was fiddling with my settings for the linear open scan, variable. More hits. Also I changed my Exec Controls from NORMAL to RELEARN. Deviation is 000.
    After changing back to NORMAL, scan is in tolerance, deviation of .00013.

    This is too good to be true, isn't it? I was under the impression that RELEARN makes the program learn the vectors during each execution. Did I accidentally reset my nominals when I did this? Is that even possible when using a model? Getting more confused...
  • Print is master, but angles are not defined, so I have to use the model for that. After zooming in close, it looks my keyed points are not actually on the model, as you suggest, but scanned points are. Not sure where I went wrong here. After starting a brand new program I have gotten my scans to agree with keyed in points (or at least much closer).



    I did to manual and DCC. Part was not touched during this process. I kept re-measuring and got the same divergent results.
    Probe Comp is ON. (If only it could be so simple...)

    YEA, IF ONLY. Is the part flat? Still the scanned values and TTP values should be very close to each other even if the part is bad.
  • The profile I'm checking is on the edge of the part, .333 thick. The edge is curved, and its cross-section has a constant .470 radius.

    I posted a partial drawing here.
  • I checked your print. Normally when I have something like that I just use Surface points. From what I can tell, What you have is Gage points, with a given XZ values. This can be very tricky, depending on your customer, some customers have a spec in which explains which of the BASICS is a set dimension and which can be toleranced. For example in the spec it may say that the distance is the SET dimesion .715, and the R .171 can be dimensioned +/-.00025. If it doesnt, than it can moved in both direction as long as it is with in .0005.

    1. When I do Gage points with a given profile of .0005, I dimension my gage pint X and Z +/-.00025.
    2. The reason why i use surface points with 3 sample hits is to assure the Vector being normal to the surface, regardless of the angle (in your case).
    3. if I see that the actuals of the surface points in one Axis is off by a little, I move around THE TARGET values in both the axis, by maybe a .0001 and see if that helps


    I could be wrong in all this, but this is what I usually do.
  • I have only just taken a look at this post and I have a number of concerns.

    The first is the datum setup on the drawing it is very confusing and in ISO land ilegal, datum ' B ' is about the only one that could be taken as correct, 'A' and 'C' are a joke. You need to have the engineer/customer state what feature are to be used for the datum setup before you can obtain any meaningful results.

    You have stated that the part has a curved edge with a constant 0.470 radius along it. From looking at the posted drawing that is not what I see.

    The part starts as a cylinder of 0.333 for a length of 0.652" ( with 2 angled flats 11.000 deg ) then transitions into constantly changing radius curve ( or is it a set of tapper cones hard to tell from drawing )
    the only place where the radius is 0.470 is at the parallel section A - A.

    You have a CAD model was this generated in-house or supplied by the customer as it might help in working out what is required.
  • I have only just taken a look at this post and I have a number of concerns.

    The first is the datum setup on the drawing it is very confusing and in ISO land ilegal, datum ' B ' is about the only one that could be taken as correct, 'A' and 'C' are a joke. You need to have the engineer/customer state what feature are to be used for the datum setup before you can obtain any meaningful results.

    You have stated that the part has a curved edge with a constant 0.470 radius along it. From looking at the posted drawing that is not what I see.

    The part starts as a cylinder of 0.333 for a length of 0.652" ( with 2 angled flats 11.000 deg ) then transitions into constantly changing radius curve ( or is it a set of tapper cones hard to tell from drawing )
    the only place where the radius is 0.470 is at the parallel section A - A.

    You have a CAD model was this generated in-house or supplied by the customer as it might help in working out what is required.
    if you look to the right, you will find a cross section view A-A with s radius of .470
  • The profile I'm checking is on the edge of the part, .333 thick. The edge is curved, and its cross-section has a constant .470 radius.

    I posted a partial drawing here.
    when you take a hit on the .652 and .333 what are your actuals?

    The .470R you can check as a radius
© 2024 Hexagon AB and/or its subsidiaries. Privacy Policy | Cloud Services Agreement