Hello,
first, let me say that I found many of the threads I've read so far very insightful and valuable.
My question concerns a methodology I should use to validate optical inspection software. The software uses structured light to take multiple images of a 2D part that are combined into a single inspection image used to extrapolate outside perimeter which is then compared to CAD.
If the extrapolated perimeter is within a set tolerance then the part passes inspection, otherwise it fails. The inspections are sensitive to the position and orientation of the part within the camera field of view. Tighter the tolerance, the worst the performance...
What I want to be able to say is that at +/- X inches the system is capable of Y sigma performance. I have 10 CMM'd parts that I can use as masters for this validation. I can also alter the associated CAD to make the system fail the part eg. I can shrink or stretch the part in CAD and then inspect it against the CMM master where it would obviously fail. I can also rotate the part within the inspection window at 45 degrees and/or place it in the middle or in the corner.
I considered using Gage R&R, I'm just not sure if that is the right method as for each inspection I only get a binary PASS or FAIL per the expected result.
Thank you and I appreciate all of your input.
Pete
first, let me say that I found many of the threads I've read so far very insightful and valuable.
My question concerns a methodology I should use to validate optical inspection software. The software uses structured light to take multiple images of a 2D part that are combined into a single inspection image used to extrapolate outside perimeter which is then compared to CAD.
If the extrapolated perimeter is within a set tolerance then the part passes inspection, otherwise it fails. The inspections are sensitive to the position and orientation of the part within the camera field of view. Tighter the tolerance, the worst the performance...
What I want to be able to say is that at +/- X inches the system is capable of Y sigma performance. I have 10 CMM'd parts that I can use as masters for this validation. I can also alter the associated CAD to make the system fail the part eg. I can shrink or stretch the part in CAD and then inspect it against the CMM master where it would obviously fail. I can also rotate the part within the inspection window at 45 degrees and/or place it in the middle or in the corner.
I considered using Gage R&R, I'm just not sure if that is the right method as for each inspection I only get a binary PASS or FAIL per the expected result.
Thank you and I appreciate all of your input.
Pete
